x86: make io_apic_32.c the same as io_apic_64.c
authorIngo Molnar <mingo@elte.hu>
Wed, 20 Aug 2008 07:07:45 +0000 (09:07 +0200)
committerIngo Molnar <mingo@elte.hu>
Thu, 16 Oct 2008 14:52:55 +0000 (16:52 +0200)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
arch/x86/kernel/io_apic_32.c

index 3ed36041c81ec93905fd05386a3f9f7d96d13bba..fba6d6ee34809e1bf7c72dbcda3f51aecaac84d1 100644 (file)
@@ -35,7 +35,7 @@
 #include <linux/htirq.h>
 #include <linux/freezer.h>
 #include <linux/kthread.h>
 #include <linux/htirq.h>
 #include <linux/freezer.h>
 #include <linux/kthread.h>
-#include <linux/jiffies.h>      /* time_after() */
+#include <linux/jiffies.h>     /* time_after() */
 #ifdef CONFIG_ACPI
 #include <acpi/acpi_bus.h>
 #endif
 #ifdef CONFIG_ACPI
 #include <acpi/acpi_bus.h>
 #endif
@@ -64,8 +64,8 @@
 #define __apicdebuginit(type) static type __init
 
 /*
 #define __apicdebuginit(type) static type __init
 
 /*
- *     Is the SiS APIC rmw bug present ?
- *     -1 = don't know, 0 = no, 1 = yes
+ *      Is the SiS APIC rmw bug present ?
+ *      -1 = don't know, 0 = no, 1 = yes
  */
 int sis_apic_bug = -1;
 
  */
 int sis_apic_bug = -1;
 
@@ -102,7 +102,7 @@ DECLARE_BITMAP(mp_bus_not_pci, MAX_MP_BUSSES);
 
 int skip_ioapic_setup;
 
 
 int skip_ioapic_setup;
 
-static int __init parse_noapic(char *arg)
+static int __init parse_noapic(char *str)
 {
        /* disable IO-APIC */
        disable_ioapic_setup();
 {
        /* disable IO-APIC */
        disable_ioapic_setup();
@@ -188,7 +188,7 @@ static void __init init_work(void *data)
        irq_cfgx[legacy_count - 1].next = NULL;
 }
 
        irq_cfgx[legacy_count - 1].next = NULL;
 }
 
-#define for_each_irq_cfg(cfg)           \
+#define for_each_irq_cfg(cfg)          \
        for (cfg = irq_cfgx; cfg; cfg = cfg->next)
 
 DEFINE_DYN_ARRAY(irq_cfgx, sizeof(struct irq_cfg), nr_irq_cfg, PAGE_SIZE, init_work);
        for (cfg = irq_cfgx; cfg; cfg = cfg->next)
 
 DEFINE_DYN_ARRAY(irq_cfgx, sizeof(struct irq_cfg), nr_irq_cfg, PAGE_SIZE, init_work);
@@ -262,7 +262,6 @@ static struct irq_cfg *irq_cfg_alloc(unsigned int irq)
                irq_cfgx = cfg;
        cfg->irq = irq;
        printk(KERN_DEBUG "found new irq_cfg for irq %d\n", cfg->irq);
                irq_cfgx = cfg;
        cfg->irq = irq;
        printk(KERN_DEBUG "found new irq_cfg for irq %d\n", cfg->irq);
-
 #ifdef CONFIG_HAVE_SPARSE_IRQ_DEBUG
        {
                /* dump the results */
 #ifdef CONFIG_HAVE_SPARSE_IRQ_DEBUG
        {
                /* dump the results */
@@ -384,9 +383,9 @@ static inline void io_apic_write(unsigned int apic, unsigned int reg, unsigned i
  */
 static inline void io_apic_modify(unsigned int apic, unsigned int reg, unsigned int value)
 {
  */
 static inline void io_apic_modify(unsigned int apic, unsigned int reg, unsigned int value)
 {
-       volatile struct io_apic __iomem *io_apic = io_apic_base(apic);
-       if (sis_apic_bug)
-               writel(reg, &io_apic->index);
+       struct io_apic __iomem *io_apic = io_apic_base(apic);
+        if (sis_apic_bug)
+                writel(reg, &io_apic->index);
        writel(value, &io_apic->data);
 }
 
        writel(value, &io_apic->data);
 }
 
@@ -494,11 +493,20 @@ static void __target_IO_APIC_irq(unsigned int irq, unsigned int dest, u8 vector)
 
                apic = entry->apic;
                pin = entry->pin;
 
                apic = entry->apic;
                pin = entry->pin;
+#ifdef CONFIG_INTR_REMAP
+               /*
+                * With interrupt-remapping, destination information comes
+                * from interrupt-remapping table entry.
+                */
+               if (!irq_remapped(irq))
+                       io_apic_write(apic, 0x11 + pin*2, dest);
+#else
                io_apic_write(apic, 0x11 + pin*2, dest);
                io_apic_write(apic, 0x11 + pin*2, dest);
+#endif
                reg = io_apic_read(apic, 0x10 + pin*2);
                reg &= ~IO_APIC_REDIR_VECTOR_MASK;
                reg |= vector;
                reg = io_apic_read(apic, 0x10 + pin*2);
                reg &= ~IO_APIC_REDIR_VECTOR_MASK;
                reg |= vector;
-               io_apic_modify(apic, 0x10 + pin *2, reg);
+               io_apic_modify(apic, 0x10 + pin*2, reg);
                if (!entry->next)
                        break;
                entry = entry->next;
                if (!entry->next)
                        break;
                entry = entry->next;
@@ -513,6 +521,7 @@ static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask)
        unsigned long flags;
        unsigned int dest;
        cpumask_t tmp;
        unsigned long flags;
        unsigned int dest;
        cpumask_t tmp;
+       struct irq_desc *desc;
 
        cpus_and(tmp, mask, cpu_online_map);
        if (cpus_empty(tmp))
 
        cpus_and(tmp, mask, cpu_online_map);
        if (cpus_empty(tmp))
@@ -529,12 +538,12 @@ static void set_ioapic_affinity_irq(unsigned int irq, cpumask_t mask)
         */
        dest = SET_APIC_LOGICAL_ID(dest);
 
         */
        dest = SET_APIC_LOGICAL_ID(dest);
 
+       desc = irq_to_desc(irq);
        spin_lock_irqsave(&ioapic_lock, flags);
        __target_IO_APIC_irq(irq, dest, cfg->vector);
        spin_lock_irqsave(&ioapic_lock, flags);
        __target_IO_APIC_irq(irq, dest, cfg->vector);
-       irq_to_desc(irq)->affinity = mask;
+       desc->affinity = mask;
        spin_unlock_irqrestore(&ioapic_lock, flags);
 }
        spin_unlock_irqrestore(&ioapic_lock, flags);
 }
-
 #endif /* CONFIG_SMP */
 
 /*
 #endif /* CONFIG_SMP */
 
 /*
@@ -699,7 +708,7 @@ static void __unmask_and_level_IO_APIC_irq(unsigned int irq)
 
 #endif
 
 
 #endif
 
-static void mask_IO_APIC_irq(unsigned int irq)
+static void mask_IO_APIC_irq (unsigned int irq)
 {
        unsigned long flags;
 
 {
        unsigned long flags;
 
@@ -708,7 +717,7 @@ static void mask_IO_APIC_irq(unsigned int irq)
        spin_unlock_irqrestore(&ioapic_lock, flags);
 }
 
        spin_unlock_irqrestore(&ioapic_lock, flags);
 }
 
-static void unmask_IO_APIC_irq(unsigned int irq)
+static void unmask_IO_APIC_irq (unsigned int irq)
 {
        unsigned long flags;
 
 {
        unsigned long flags;
 
@@ -725,14 +734,13 @@ static void clear_IO_APIC_pin(unsigned int apic, unsigned int pin)
        entry = ioapic_read_entry(apic, pin);
        if (entry.delivery_mode == dest_SMI)
                return;
        entry = ioapic_read_entry(apic, pin);
        if (entry.delivery_mode == dest_SMI)
                return;
-
        /*
         * Disable it in the IO-APIC irq-routing table:
         */
        ioapic_mask_entry(apic, pin);
 }
 
        /*
         * Disable it in the IO-APIC irq-routing table:
         */
        ioapic_mask_entry(apic, pin);
 }
 
-static void clear_IO_APIC(void)
+static void clear_IO_APIC (void)
 {
        int apic, pin;
 
 {
        int apic, pin;
 
@@ -741,7 +749,7 @@ static void clear_IO_APIC(void)
                        clear_IO_APIC_pin(apic, pin);
 }
 
                        clear_IO_APIC_pin(apic, pin);
 }
 
-#ifndef CONFIG_SMP
+#if !defined(CONFIG_SMP) && defined(CONFIG_X86_32)
 void send_IPI_self(int vector)
 {
        unsigned int cfg;
 void send_IPI_self(int vector)
 {
        unsigned int cfg;
@@ -756,9 +764,9 @@ void send_IPI_self(int vector)
         */
        apic_write(APIC_ICR, cfg);
 }
         */
        apic_write(APIC_ICR, cfg);
 }
-#endif /* !CONFIG_SMP */
-
+#endif /* !CONFIG_SMP && CONFIG_X86_32*/
 
 
+#ifdef CONFIG_X86_32
 /*
  * support for broken MP BIOSs, enables hand-redirection of PIRQ0-7 to
  * specific CPU-side IRQs.
 /*
  * support for broken MP BIOSs, enables hand-redirection of PIRQ0-7 to
  * specific CPU-side IRQs.
@@ -797,6 +805,75 @@ static int __init ioapic_pirq_setup(char *str)
 }
 
 __setup("pirq=", ioapic_pirq_setup);
 }
 
 __setup("pirq=", ioapic_pirq_setup);
+#endif /* CONFIG_X86_32 */
+
+#ifdef CONFIG_INTR_REMAP
+/* I/O APIC RTE contents at the OS boot up */
+static struct IO_APIC_route_entry *early_ioapic_entries[MAX_IO_APICS];
+
+/*
+ * Saves and masks all the unmasked IO-APIC RTE's
+ */
+int save_mask_IO_APIC_setup(void)
+{
+       union IO_APIC_reg_01 reg_01;
+       unsigned long flags;
+       int apic, pin;
+
+       /*
+        * The number of IO-APIC IRQ registers (== #pins):
+        */
+       for (apic = 0; apic < nr_ioapics; apic++) {
+               spin_lock_irqsave(&ioapic_lock, flags);
+               reg_01.raw = io_apic_read(apic, 1);
+               spin_unlock_irqrestore(&ioapic_lock, flags);
+               nr_ioapic_registers[apic] = reg_01.bits.entries+1;
+       }
+
+       for (apic = 0; apic < nr_ioapics; apic++) {
+               early_ioapic_entries[apic] =
+                       kzalloc(sizeof(struct IO_APIC_route_entry) *
+                               nr_ioapic_registers[apic], GFP_KERNEL);
+               if (!early_ioapic_entries[apic])
+                       return -ENOMEM;
+       }
+
+       for (apic = 0; apic < nr_ioapics; apic++)
+               for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
+                       struct IO_APIC_route_entry entry;
+
+                       entry = early_ioapic_entries[apic][pin] =
+                               ioapic_read_entry(apic, pin);
+                       if (!entry.mask) {
+                               entry.mask = 1;
+                               ioapic_write_entry(apic, pin, entry);
+                       }
+               }
+       return 0;
+}
+
+void restore_IO_APIC_setup(void)
+{
+       int apic, pin;
+
+       for (apic = 0; apic < nr_ioapics; apic++)
+               for (pin = 0; pin < nr_ioapic_registers[apic]; pin++)
+                       ioapic_write_entry(apic, pin,
+                                          early_ioapic_entries[apic][pin]);
+}
+
+void reinit_intr_remapped_IO_APIC(int intr_remapping)
+{
+       /*
+        * for now plain restore of previous settings.
+        * TBD: In the case of OS enabling interrupt-remapping,
+        * IO-APIC RTE's need to be setup to point to interrupt-remapping
+        * table entries. for now, do a plain restore, and wait for
+        * the setup_IO_APIC_irqs() to do proper initialization.
+        */
+       restore_IO_APIC_setup();
+}
+#endif
 
 /*
  * Find the IRQ entry number of a certain pin.
 
 /*
  * Find the IRQ entry number of a certain pin.
@@ -848,7 +925,7 @@ static int __init find_isa_irq_apic(int irq, int type)
        }
        if (i < mp_irq_entries) {
                int apic;
        }
        if (i < mp_irq_entries) {
                int apic;
-               for (apic = 0; apic < nr_ioapics; apic++) {
+               for(apic = 0; apic < nr_ioapics; apic++) {
                        if (mp_ioapics[apic].mp_apicid == mp_irqs[i].mp_dstapic)
                                return apic;
                }
                        if (mp_ioapics[apic].mp_apicid == mp_irqs[i].mp_dstapic)
                                return apic;
                }
@@ -867,10 +944,10 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin)
 {
        int apic, i, best_guess = -1;
 
 {
        int apic, i, best_guess = -1;
 
-       apic_printk(APIC_DEBUG, "querying PCI -> IRQ mapping bus:%d, "
-               "slot:%d, pin:%d.\n", bus, slot, pin);
+       apic_printk(APIC_DEBUG, "querying PCI -> IRQ mapping bus:%d, slot:%d, pin:%d.\n",
+               bus, slot, pin);
        if (test_bit(bus, mp_bus_not_pci)) {
        if (test_bit(bus, mp_bus_not_pci)) {
-               printk(KERN_WARNING "PCI BIOS passed nonexistent PCI bus %d!\n", bus);
+               apic_printk(APIC_VERBOSE, "PCI BIOS passed nonexistent PCI bus %d!\n", bus);
                return -1;
        }
        for (i = 0; i < mp_irq_entries; i++) {
                return -1;
        }
        for (i = 0; i < mp_irq_entries; i++) {
@@ -885,7 +962,7 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin)
                    !mp_irqs[i].mp_irqtype &&
                    (bus == lbus) &&
                    (slot == ((mp_irqs[i].mp_srcbusirq >> 2) & 0x1f))) {
                    !mp_irqs[i].mp_irqtype &&
                    (bus == lbus) &&
                    (slot == ((mp_irqs[i].mp_srcbusirq >> 2) & 0x1f))) {
-                       int irq = pin_2_irq(i, apic, mp_irqs[i].mp_dstirq);
+                       int irq = pin_2_irq(i,apic,mp_irqs[i].mp_dstirq);
 
                        if (!(apic || IO_APIC_IRQ(irq)))
                                continue;
 
                        if (!(apic || IO_APIC_IRQ(irq)))
                                continue;
@@ -902,6 +979,7 @@ int IO_APIC_get_PCI_irq_vector(int bus, int slot, int pin)
        }
        return best_guess;
 }
        }
        return best_guess;
 }
+
 EXPORT_SYMBOL(IO_APIC_get_PCI_irq_vector);
 
 #if defined(CONFIG_EISA) || defined(CONFIG_MCA)
 EXPORT_SYMBOL(IO_APIC_get_PCI_irq_vector);
 
 #if defined(CONFIG_EISA) || defined(CONFIG_MCA)
@@ -918,6 +996,7 @@ static int EISA_ELCR(unsigned int irq)
                        "Broken MPtable reports ISA irq %d\n", irq);
        return 0;
 }
                        "Broken MPtable reports ISA irq %d\n", irq);
        return 0;
 }
+
 #endif
 
 /* ISA interrupts are always polarity zero edge triggered,
 #endif
 
 /* ISA interrupts are always polarity zero edge triggered,
@@ -954,36 +1033,36 @@ static int MPBIOS_polarity(int idx)
        /*
         * Determine IRQ line polarity (high active or low active):
         */
        /*
         * Determine IRQ line polarity (high active or low active):
         */
-       switch (mp_irqs[idx].mp_irqflag & 3) {
-       case 0: /* conforms, ie. bus-type dependent polarity */
-       {
-               polarity = test_bit(bus, mp_bus_not_pci)?
-                       default_ISA_polarity(idx):
-                       default_PCI_polarity(idx);
-               break;
-       }
-       case 1: /* high active */
-       {
-               polarity = 0;
-               break;
-       }
-       case 2: /* reserved */
-       {
-               printk(KERN_WARNING "broken BIOS!!\n");
-               polarity = 1;
-               break;
-       }
-       case 3: /* low active */
-       {
-               polarity = 1;
-               break;
-       }
-       default: /* invalid */
+       switch (mp_irqs[idx].mp_irqflag & 3)
        {
        {
-               printk(KERN_WARNING "broken BIOS!!\n");
-               polarity = 1;
-               break;
-       }
+               case 0: /* conforms, ie. bus-type dependent polarity */
+                       if (test_bit(bus, mp_bus_not_pci))
+                               polarity = default_ISA_polarity(idx);
+                       else
+                               polarity = default_PCI_polarity(idx);
+                       break;
+               case 1: /* high active */
+               {
+                       polarity = 0;
+                       break;
+               }
+               case 2: /* reserved */
+               {
+                       printk(KERN_WARNING "broken BIOS!!\n");
+                       polarity = 1;
+                       break;
+               }
+               case 3: /* low active */
+               {
+                       polarity = 1;
+                       break;
+               }
+               default: /* invalid */
+               {
+                       printk(KERN_WARNING "broken BIOS!!\n");
+                       polarity = 1;
+                       break;
+               }
        }
        return polarity;
 }
        }
        return polarity;
 }
@@ -996,67 +1075,67 @@ static int MPBIOS_trigger(int idx)
        /*
         * Determine IRQ trigger mode (edge or level sensitive):
         */
        /*
         * Determine IRQ trigger mode (edge or level sensitive):
         */
-       switch ((mp_irqs[idx].mp_irqflag>>2) & 3) {
-       case 0: /* conforms, ie. bus-type dependent */
+       switch ((mp_irqs[idx].mp_irqflag>>2) & 3)
        {
        {
-               trigger = test_bit(bus, mp_bus_not_pci)?
-                               default_ISA_trigger(idx):
-                               default_PCI_trigger(idx);
+               case 0: /* conforms, ie. bus-type dependent */
+                       if (test_bit(bus, mp_bus_not_pci))
+                               trigger = default_ISA_trigger(idx);
+                       else
+                               trigger = default_PCI_trigger(idx);
 #if defined(CONFIG_EISA) || defined(CONFIG_MCA)
 #if defined(CONFIG_EISA) || defined(CONFIG_MCA)
-               switch (mp_bus_id_to_type[bus]) {
-               case MP_BUS_ISA: /* ISA pin */
-               {
-                       /* set before the switch */
+                       switch (mp_bus_id_to_type[bus]) {
+                               case MP_BUS_ISA: /* ISA pin */
+                               {
+                                       /* set before the switch */
+                                       break;
+                               }
+                               case MP_BUS_EISA: /* EISA pin */
+                               {
+                                       trigger = default_EISA_trigger(idx);
+                                       break;
+                               }
+                               case MP_BUS_PCI: /* PCI pin */
+                               {
+                                       /* set before the switch */
+                                       break;
+                               }
+                               case MP_BUS_MCA: /* MCA pin */
+                               {
+                                       trigger = default_MCA_trigger(idx);
+                                       break;
+                               }
+                               default:
+                               {
+                                       printk(KERN_WARNING "broken BIOS!!\n");
+                                       trigger = 1;
+                                       break;
+                               }
+                       }
+#endif
                        break;
                        break;
-               }
-               case MP_BUS_EISA: /* EISA pin */
+               case 1: /* edge */
                {
                {
-                       trigger = default_EISA_trigger(idx);
+                       trigger = 0;
                        break;
                }
                        break;
                }
-               case MP_BUS_PCI: /* PCI pin */
+               case 2: /* reserved */
                {
                {
-                       /* set before the switch */
+                       printk(KERN_WARNING "broken BIOS!!\n");
+                       trigger = 1;
                        break;
                }
                        break;
                }
-               case MP_BUS_MCA: /* MCA pin */
+               case 3: /* level */
                {
                {
-                       trigger = default_MCA_trigger(idx);
+                       trigger = 1;
                        break;
                }
                        break;
                }
-               default:
+               default: /* invalid */
                {
                        printk(KERN_WARNING "broken BIOS!!\n");
                {
                        printk(KERN_WARNING "broken BIOS!!\n");
-                       trigger = 1;
+                       trigger = 0;
                        break;
                }
        }
                        break;
                }
        }
-#endif
-               break;
-       }
-       case 1: /* edge */
-       {
-               trigger = 0;
-               break;
-       }
-       case 2: /* reserved */
-       {
-               printk(KERN_WARNING "broken BIOS!!\n");
-               trigger = 1;
-               break;
-       }
-       case 3: /* level */
-       {
-               trigger = 1;
-               break;
-       }
-       default: /* invalid */
-       {
-               printk(KERN_WARNING "broken BIOS!!\n");
-               trigger = 0;
-               break;
-       }
-       }
        return trigger;
 }
 
        return trigger;
 }
 
@@ -1082,9 +1161,9 @@ static int pin_2_irq(int idx, int apic, int pin)
        if (mp_irqs[idx].mp_dstirq != pin)
                printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n");
 
        if (mp_irqs[idx].mp_dstirq != pin)
                printk(KERN_ERR "broken BIOS or MPTABLE parser, ayiee!!\n");
 
-       if (test_bit(bus, mp_bus_not_pci))
+       if (test_bit(bus, mp_bus_not_pci)) {
                irq = mp_irqs[idx].mp_srcbusirq;
                irq = mp_irqs[idx].mp_srcbusirq;
-       else {
+       else {
                /*
                 * PCI IRQs are mapped in order
                 */
                /*
                 * PCI IRQs are mapped in order
                 */
@@ -1092,14 +1171,14 @@ static int pin_2_irq(int idx, int apic, int pin)
                while (i < apic)
                        irq += nr_ioapic_registers[i++];
                irq += pin;
                while (i < apic)
                        irq += nr_ioapic_registers[i++];
                irq += pin;
-
-               /*
-                * For MPS mode, so far only needed by ES7000 platform
-                */
-               if (ioapic_renumber_irq)
-                       irq = ioapic_renumber_irq(apic, irq);
+                /*
+                 * For MPS mode, so far only needed by ES7000 platform
+                 */
+                if (ioapic_renumber_irq)
+                        irq = ioapic_renumber_irq(apic, irq);
        }
 
        }
 
+#ifdef CONFIG_X86_32
        /*
         * PCI IRQ command line redirection. Yes, limits are hardcoded.
         */
        /*
         * PCI IRQ command line redirection. Yes, limits are hardcoded.
         */
@@ -1116,6 +1195,8 @@ static int pin_2_irq(int idx, int apic, int pin)
                        }
                }
        }
                        }
                }
        }
+#endif
+
        return irq;
 }
 
        return irq;
 }
 
@@ -1145,74 +1226,70 @@ static int __assign_irq_vector(int irq, cpumask_t mask)
         * Also, we've got to be careful not to trash gate
         * 0x80, because int 0x80 is hm, kind of importantish. ;)
         */
         * Also, we've got to be careful not to trash gate
         * 0x80, because int 0x80 is hm, kind of importantish. ;)
         */
-        static int current_vector = FIRST_DEVICE_VECTOR, current_offset = 0;
-        unsigned int old_vector;
-        int cpu;
-        struct irq_cfg *cfg;
+       static int current_vector = FIRST_DEVICE_VECTOR, current_offset = 0;
+       unsigned int old_vector;
+       int cpu;
+       struct irq_cfg *cfg;
 
 
-        cfg = irq_cfg(irq);
+       cfg = irq_cfg(irq);
 
 
-        /* Only try and allocate irqs on cpus that are present */
-        cpus_and(mask, mask, cpu_online_map);
+       /* Only try and allocate irqs on cpus that are present */
+       cpus_and(mask, mask, cpu_online_map);
 
 
-        if ((cfg->move_in_progress) || cfg->move_cleanup_count)
-                return -EBUSY;
+       if ((cfg->move_in_progress) || cfg->move_cleanup_count)
+               return -EBUSY;
 
 
-        old_vector = cfg->vector;
-        if (old_vector) {
-                cpumask_t tmp;
-                cpus_and(tmp, cfg->domain, mask);
-                if (!cpus_empty(tmp))
-                        return 0;
-        }
+       old_vector = cfg->vector;
+       if (old_vector) {
+               cpumask_t tmp;
+               cpus_and(tmp, cfg->domain, mask);
+               if (!cpus_empty(tmp))
+                       return 0;
+       }
 
 
-        for_each_cpu_mask_nr(cpu, mask) {
-                cpumask_t domain, new_mask;
-                int new_cpu;
-                int vector, offset;
+       for_each_cpu_mask_nr(cpu, mask) {
+               cpumask_t domain, new_mask;
+               int new_cpu;
+               int vector, offset;
 
 
-                domain = vector_allocation_domain(cpu);
-                cpus_and(new_mask, domain, cpu_online_map);
+               domain = vector_allocation_domain(cpu);
+               cpus_and(new_mask, domain, cpu_online_map);
 
 
-                vector = current_vector;
-                offset = current_offset;
+               vector = current_vector;
+               offset = current_offset;
 next:
 next:
-                vector += 8;
-                if (vector >= first_system_vector) {
-                        /* If we run out of vectors on large boxen, must share them. */
-                        offset = (offset + 1) % 8;
-                        vector = FIRST_DEVICE_VECTOR + offset;
-                }
-                if (unlikely(current_vector == vector))
-                        continue;
+               vector += 8;
+               if (vector >= first_system_vector) {
+                       /* If we run out of vectors on large boxen, must share them. */
+                       offset = (offset + 1) % 8;
+                       vector = FIRST_DEVICE_VECTOR + offset;
+               }
+               if (unlikely(current_vector == vector))
+                       continue;
 #ifdef CONFIG_X86_64
 #ifdef CONFIG_X86_64
-                if (vector == IA32_SYSCALL_VECTOR)
-                        goto next;
+               if (vector == IA32_SYSCALL_VECTOR)
+                       goto next;
 #else
 #else
-                if (vector == SYSCALL_VECTOR)
-                        goto next;
+               if (vector == SYSCALL_VECTOR)
+                       goto next;
 #endif
 #endif
-                for_each_cpu_mask_nr(new_cpu, new_mask)
-                        if (per_cpu(vector_irq, new_cpu)[vector] != -1)
-                                goto next;
-                /* Found one! */
-                current_vector = vector;
-                current_offset = offset;
-                if (old_vector) {
-                        cfg->move_in_progress = 1;
-                        cfg->old_domain = cfg->domain;
-                }
-               printk(KERN_DEBUG "assign_irq_vector: irq %d vector %#x cpu ", irq, vector);
-               for_each_cpu_mask_nr(new_cpu, new_mask) {
-                       per_cpu(vector_irq, new_cpu)[vector] = irq;
-                       printk(KERN_CONT " %d ", new_cpu);
+               for_each_cpu_mask_nr(new_cpu, new_mask)
+                       if (per_cpu(vector_irq, new_cpu)[vector] != -1)
+                               goto next;
+               /* Found one! */
+               current_vector = vector;
+               current_offset = offset;
+               if (old_vector) {
+                       cfg->move_in_progress = 1;
+                       cfg->old_domain = cfg->domain;
                }
                }
-               printk(KERN_CONT "\n");
-                cfg->vector = vector;
-                cfg->domain = domain;
-                return 0;
-        }
-        return -ENOSPC;
+               for_each_cpu_mask_nr(new_cpu, new_mask)
+                       per_cpu(vector_irq, new_cpu)[vector] = irq;
+               cfg->vector = vector;
+               cfg->domain = domain;
+               return 0;
+       }
+       return -ENOSPC;
 }
 
 static int assign_irq_vector(int irq, cpumask_t mask)
 }
 
 static int assign_irq_vector(int irq, cpumask_t mask)
@@ -1223,7 +1300,6 @@ static int assign_irq_vector(int irq, cpumask_t mask)
        spin_lock_irqsave(&vector_lock, flags);
        err = __assign_irq_vector(irq, mask);
        spin_unlock_irqrestore(&vector_lock, flags);
        spin_lock_irqsave(&vector_lock, flags);
        err = __assign_irq_vector(irq, mask);
        spin_unlock_irqrestore(&vector_lock, flags);
-
        return err;
 }
 
        return err;
 }
 
@@ -1269,36 +1345,39 @@ void __setup_vector_irq(int cpu)
                cfg = irq_cfg(irq);
                if (!cpu_isset(cpu, cfg->domain))
                        per_cpu(vector_irq, cpu)[vector] = -1;
                cfg = irq_cfg(irq);
                if (!cpu_isset(cpu, cfg->domain))
                        per_cpu(vector_irq, cpu)[vector] = -1;
-        }
+       }
 }
 
 static struct irq_chip ioapic_chip;
 }
 
 static struct irq_chip ioapic_chip;
+#ifdef CONFIG_INTR_REMAP
+static struct irq_chip ir_ioapic_chip;
+#endif
 
 
-#define IOAPIC_AUTO    -1
-#define IOAPIC_EDGE    0
-#define IOAPIC_LEVEL   1
+#define IOAPIC_AUTO     -1
+#define IOAPIC_EDGE     0
+#define IOAPIC_LEVEL    1
 
 #ifdef CONFIG_X86_32
 static inline int IO_APIC_irq_trigger(int irq)
 {
 
 #ifdef CONFIG_X86_32
 static inline int IO_APIC_irq_trigger(int irq)
 {
-       int apic, idx, pin;
+        int apic, idx, pin;
 
 
-       for (apic = 0; apic < nr_ioapics; apic++) {
-               for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
-                       idx = find_irq_entry(apic, pin, mp_INT);
-                       if ((idx != -1) && (irq == pin_2_irq(idx, apic, pin)))
-                               return irq_trigger(idx);
-               }
-       }
-       /*
-        * nonexistent IRQs are edge default
-        */
-       return 0;
+        for (apic = 0; apic < nr_ioapics; apic++) {
+                for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
+                        idx = find_irq_entry(apic, pin, mp_INT);
+                        if ((idx != -1) && (irq == pin_2_irq(idx, apic, pin)))
+                                return irq_trigger(idx);
+                }
+        }
+        /*
+         * nonexistent IRQs are edge default
+         */
+        return 0;
 }
 #else
 static inline int IO_APIC_irq_trigger(int irq)
 {
 }
 #else
 static inline int IO_APIC_irq_trigger(int irq)
 {
-        return 1;
+       return 1;
 }
 #endif
 
 }
 #endif
 
@@ -1318,13 +1397,27 @@ static void ioapic_register_intr(int irq, unsigned long trigger)
        else
                desc->status &= ~IRQ_LEVEL;
 
        else
                desc->status &= ~IRQ_LEVEL;
 
+#ifdef CONFIG_INTR_REMAP
+       if (irq_remapped(irq)) {
+               desc->status |= IRQ_MOVE_PCNTXT;
+               if (trigger)
+                       set_irq_chip_and_handler_name(irq, &ir_ioapic_chip,
+                                                     handle_fasteoi_irq,
+                                                    "fasteoi");
+               else
+                       set_irq_chip_and_handler_name(irq, &ir_ioapic_chip,
+                                                     handle_edge_irq, "edge");
+               return;
+       }
+#endif
        if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) ||
            trigger == IOAPIC_LEVEL)
                set_irq_chip_and_handler_name(irq, &ioapic_chip,
        if ((trigger == IOAPIC_AUTO && IO_APIC_irq_trigger(irq)) ||
            trigger == IOAPIC_LEVEL)
                set_irq_chip_and_handler_name(irq, &ioapic_chip,
-                                        handle_fasteoi_irq, "fasteoi");
+                                             handle_fasteoi_irq,
+                                             "fasteoi");
        else
                set_irq_chip_and_handler_name(irq, &ioapic_chip,
        else
                set_irq_chip_and_handler_name(irq, &ioapic_chip,
-                                        handle_edge_irq, "edge");
+                                             handle_edge_irq, "edge");
 }
 
 static int setup_ioapic_entry(int apic, int irq,
 }
 
 static int setup_ioapic_entry(int apic, int irq,
@@ -1337,11 +1430,45 @@ static int setup_ioapic_entry(int apic, int irq,
         */
        memset(entry,0,sizeof(*entry));
 
         */
        memset(entry,0,sizeof(*entry));
 
-       entry->delivery_mode = INT_DELIVERY_MODE;
-       entry->dest_mode = INT_DEST_MODE;
-       entry->dest = destination;
+#ifdef CONFIG_INTR_REMAP
+       if (intr_remapping_enabled) {
+               struct intel_iommu *iommu = map_ioapic_to_ir(apic);
+               struct irte irte;
+               struct IR_IO_APIC_route_entry *ir_entry =
+                       (struct IR_IO_APIC_route_entry *) entry;
+               int index;
+
+               if (!iommu)
+                       panic("No mapping iommu for ioapic %d\n", apic);
+
+               index = alloc_irte(iommu, irq, 1);
+               if (index < 0)
+                       panic("Failed to allocate IRTE for ioapic %d\n", apic);
+
+               memset(&irte, 0, sizeof(irte));
+
+               irte.present = 1;
+               irte.dst_mode = INT_DEST_MODE;
+               irte.trigger_mode = trigger;
+               irte.dlvry_mode = INT_DELIVERY_MODE;
+               irte.vector = vector;
+               irte.dest_id = IRTE_DEST(destination);
+
+               modify_irte(irq, &irte);
+
+               ir_entry->index2 = (index >> 15) & 0x1;
+               ir_entry->zero = 0;
+               ir_entry->format = 1;
+               ir_entry->index = (index & 0x7fff);
+       } else
+#endif
+       {
+               entry->delivery_mode = INT_DELIVERY_MODE;
+               entry->dest_mode = INT_DEST_MODE;
+               entry->dest = destination;
+       }
 
 
-       entry->mask = 0;                                /* enable IRQ */
+       entry->mask = 0;                                /* enable IRQ */
        entry->trigger = trigger;
        entry->polarity = polarity;
        entry->vector = vector;
        entry->trigger = trigger;
        entry->polarity = polarity;
        entry->vector = vector;
@@ -1351,12 +1478,11 @@ static int setup_ioapic_entry(int apic, int irq,
         */
        if (trigger)
                entry->mask = 1;
         */
        if (trigger)
                entry->mask = 1;
-
        return 0;
 }
 
 static void setup_IO_APIC_irq(int apic, int pin, unsigned int irq,
        return 0;
 }
 
 static void setup_IO_APIC_irq(int apic, int pin, unsigned int irq,
-                              int trigger, int polarity)
+                             int trigger, int polarity)
 {
        struct irq_cfg *cfg;
        struct IO_APIC_route_entry entry;
 {
        struct irq_cfg *cfg;
        struct IO_APIC_route_entry entry;
@@ -1420,10 +1546,10 @@ static void __init setup_IO_APIC_irqs(void)
                }
 
                irq = pin_2_irq(idx, apic, pin);
                }
 
                irq = pin_2_irq(idx, apic, pin);
-
+#ifdef CONFIG_X86_32
                 if (multi_timer_check(apic, irq))
                         continue;
                 if (multi_timer_check(apic, irq))
                         continue;
-
+#endif
                add_pin_to_irq(irq, apic, pin);
 
                setup_IO_APIC_irq(apic, pin, irq,
                add_pin_to_irq(irq, apic, pin);
 
                setup_IO_APIC_irq(apic, pin, irq,
@@ -1443,6 +1569,11 @@ static void __init setup_timer_IRQ0_pin(unsigned int apic, unsigned int pin,
 {
        struct IO_APIC_route_entry entry;
 
 {
        struct IO_APIC_route_entry entry;
 
+#ifdef CONFIG_INTR_REMAP
+       if (intr_remapping_enabled)
+               return;
+#endif
+
        memset(&entry, 0, sizeof(entry));
 
        /*
        memset(&entry, 0, sizeof(entry));
 
        /*
@@ -1461,7 +1592,7 @@ static void __init setup_timer_IRQ0_pin(unsigned int apic, unsigned int pin,
         * The timer IRQ doesn't have to know that behind the
         * scene we may have a 8259A-master in AEOI mode ...
         */
         * The timer IRQ doesn't have to know that behind the
         * scene we may have a 8259A-master in AEOI mode ...
         */
-       ioapic_register_intr(0, IOAPIC_EDGE);
+       set_irq_chip_and_handler_name(0, &ioapic_chip, handle_edge_irq, "edge");
 
        /*
         * Add it to the IO-APIC irq-routing table:
 
        /*
         * Add it to the IO-APIC irq-routing table:
@@ -1501,17 +1632,18 @@ __apicdebuginit(void) print_IO_APIC(void)
        reg_01.raw = io_apic_read(apic, 1);
        if (reg_01.bits.version >= 0x10)
                reg_02.raw = io_apic_read(apic, 2);
        reg_01.raw = io_apic_read(apic, 1);
        if (reg_01.bits.version >= 0x10)
                reg_02.raw = io_apic_read(apic, 2);
-       if (reg_01.bits.version >= 0x20)
-               reg_03.raw = io_apic_read(apic, 3);
+        if (reg_01.bits.version >= 0x20)
+                reg_03.raw = io_apic_read(apic, 3);
        spin_unlock_irqrestore(&ioapic_lock, flags);
 
        spin_unlock_irqrestore(&ioapic_lock, flags);
 
+       printk("\n");
        printk(KERN_DEBUG "IO APIC #%d......\n", mp_ioapics[apic].mp_apicid);
        printk(KERN_DEBUG ".... register #00: %08X\n", reg_00.raw);
        printk(KERN_DEBUG ".......    : physical APIC id: %02X\n", reg_00.bits.ID);
        printk(KERN_DEBUG ".......    : Delivery Type: %X\n", reg_00.bits.delivery_type);
        printk(KERN_DEBUG ".......    : LTS          : %X\n", reg_00.bits.LTS);
 
        printk(KERN_DEBUG "IO APIC #%d......\n", mp_ioapics[apic].mp_apicid);
        printk(KERN_DEBUG ".... register #00: %08X\n", reg_00.raw);
        printk(KERN_DEBUG ".......    : physical APIC id: %02X\n", reg_00.bits.ID);
        printk(KERN_DEBUG ".......    : Delivery Type: %X\n", reg_00.bits.delivery_type);
        printk(KERN_DEBUG ".......    : LTS          : %X\n", reg_00.bits.LTS);
 
-       printk(KERN_DEBUG ".... register #01: %08X\n", reg_01.raw);
+       printk(KERN_DEBUG ".... register #01: %08X\n", *(int *)&reg_01);
        printk(KERN_DEBUG ".......     : max redirection entries: %04X\n", reg_01.bits.entries);
 
        printk(KERN_DEBUG ".......     : PRQ implemented: %X\n", reg_01.bits.PRQ);
        printk(KERN_DEBUG ".......     : max redirection entries: %04X\n", reg_01.bits.entries);
 
        printk(KERN_DEBUG ".......     : PRQ implemented: %X\n", reg_01.bits.PRQ);
@@ -1548,7 +1680,10 @@ __apicdebuginit(void) print_IO_APIC(void)
 
                entry = ioapic_read_entry(apic, i);
 
 
                entry = ioapic_read_entry(apic, i);
 
-               printk(KERN_DEBUG " %02x %02X  ", i, entry.dest);
+               printk(KERN_DEBUG " %02x %03X ",
+                       i,
+                       entry.dest
+               );
 
                printk("%1d    %1d    %1d   %1d   %1d    %1d    %1d    %02X\n",
                        entry.mask,
 
                printk("%1d    %1d    %1d   %1d   %1d    %1d    %1d    %02X\n",
                        entry.mask,
@@ -1567,7 +1702,7 @@ __apicdebuginit(void) print_IO_APIC(void)
                struct irq_pin_list *entry = cfg->irq_2_pin;
                if (!entry)
                        continue;
                struct irq_pin_list *entry = cfg->irq_2_pin;
                if (!entry)
                        continue;
-               printk(KERN_DEBUG "IRQ%d ", i);
+               printk(KERN_DEBUG "IRQ%d ", cfg->irq);
                for (;;) {
                        printk("-> %d:%d", entry->apic, entry->pin);
                        if (!entry->next)
                for (;;) {
                        printk("-> %d:%d", entry->apic, entry->pin);
                        if (!entry->next)
@@ -1614,8 +1749,7 @@ __apicdebuginit(void) print_local_APIC(void *dummy)
        printk("\n" KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n",
                smp_processor_id(), hard_smp_processor_id());
        v = apic_read(APIC_ID);
        printk("\n" KERN_DEBUG "printing local APIC contents on CPU#%d/%d:\n",
                smp_processor_id(), hard_smp_processor_id());
        v = apic_read(APIC_ID);
-       printk(KERN_INFO "... APIC ID:      %08x (%01x)\n", v,
-                       GET_APIC_ID(v));
+       printk(KERN_INFO "... APIC ID:      %08x (%01x)\n", v, read_apic_id());
        v = apic_read(APIC_LVR);
        printk(KERN_INFO "... APIC VERSION: %08x\n", v);
        ver = GET_APIC_VERSION(v);
        v = apic_read(APIC_LVR);
        printk(KERN_INFO "... APIC VERSION: %08x\n", v);
        ver = GET_APIC_VERSION(v);
@@ -1624,7 +1758,7 @@ __apicdebuginit(void) print_local_APIC(void *dummy)
        v = apic_read(APIC_TASKPRI);
        printk(KERN_DEBUG "... APIC TASKPRI: %08x (%02x)\n", v, v & APIC_TPRI_MASK);
 
        v = apic_read(APIC_TASKPRI);
        printk(KERN_DEBUG "... APIC TASKPRI: %08x (%02x)\n", v, v & APIC_TPRI_MASK);
 
-       if (APIC_INTEGRATED(ver)) {                     /* !82489DX */
+       if (APIC_INTEGRATED(ver)) {                     /* !82489DX */
                v = apic_read(APIC_ARBPRI);
                printk(KERN_DEBUG "... APIC ARBPRI: %08x (%02x)\n", v,
                        v & APIC_ARBPRI_MASK);
                v = apic_read(APIC_ARBPRI);
                printk(KERN_DEBUG "... APIC ARBPRI: %08x (%02x)\n", v,
                        v & APIC_ARBPRI_MASK);
@@ -1650,9 +1784,10 @@ __apicdebuginit(void) print_local_APIC(void *dummy)
        printk(KERN_DEBUG "... APIC IRR field:\n");
        print_APIC_bitfield(APIC_IRR);
 
        printk(KERN_DEBUG "... APIC IRR field:\n");
        print_APIC_bitfield(APIC_IRR);
 
-       if (APIC_INTEGRATED(ver)) {             /* !82489DX */
-               if (maxlvt > 3)         /* Due to the Pentium erratum 3AP. */
+       if (APIC_INTEGRATED(ver)) {             /* !82489DX */
+               if (maxlvt > 3)         /* Due to the Pentium erratum 3AP. */
                        apic_write(APIC_ESR, 0);
                        apic_write(APIC_ESR, 0);
+
                v = apic_read(APIC_ESR);
                printk(KERN_DEBUG "... APIC ESR: %08x\n", v);
        }
                v = apic_read(APIC_ESR);
                printk(KERN_DEBUG "... APIC ESR: %08x\n", v);
        }
@@ -1710,11 +1845,11 @@ __apicdebuginit(void) print_PIC(void)
        v = inb(0xa0) << 8 | inb(0x20);
        printk(KERN_DEBUG "... PIC  IRR: %04x\n", v);
 
        v = inb(0xa0) << 8 | inb(0x20);
        printk(KERN_DEBUG "... PIC  IRR: %04x\n", v);
 
-       outb(0x0b, 0xa0);
-       outb(0x0b, 0x20);
+       outb(0x0b,0xa0);
+       outb(0x0b,0x20);
        v = inb(0xa0) << 8 | inb(0x20);
        v = inb(0xa0) << 8 | inb(0x20);
-       outb(0x0a, 0xa0);
-       outb(0x0a, 0x20);
+       outb(0x0a,0xa0);
+       outb(0x0a,0x20);
 
        spin_unlock_irqrestore(&i8259A_lock, flags);
 
 
        spin_unlock_irqrestore(&i8259A_lock, flags);
 
@@ -1739,16 +1874,19 @@ fs_initcall(print_all_ICs);
 /* Where if anywhere is the i8259 connect in external int mode */
 static struct { int pin, apic; } ioapic_i8259 = { -1, -1 };
 
 /* Where if anywhere is the i8259 connect in external int mode */
 static struct { int pin, apic; } ioapic_i8259 = { -1, -1 };
 
-static void __init enable_IO_APIC(void)
+void __init enable_IO_APIC(void)
 {
        union IO_APIC_reg_01 reg_01;
        int i8259_apic, i8259_pin;
 {
        union IO_APIC_reg_01 reg_01;
        int i8259_apic, i8259_pin;
-       int i, apic;
+       int apic;
        unsigned long flags;
 
        unsigned long flags;
 
+#ifdef CONFIG_X86_32
+       int i;
        if (!pirqs_enabled)
                for (i = 0; i < MAX_PIRQS; i++)
                        pirq_entries[i] = -1;
        if (!pirqs_enabled)
                for (i = 0; i < MAX_PIRQS; i++)
                        pirq_entries[i] = -1;
+#endif
 
        /*
         * The number of IO-APIC IRQ registers (== #pins):
 
        /*
         * The number of IO-APIC IRQ registers (== #pins):
@@ -1759,7 +1897,7 @@ static void __init enable_IO_APIC(void)
                spin_unlock_irqrestore(&ioapic_lock, flags);
                nr_ioapic_registers[apic] = reg_01.bits.entries+1;
        }
                spin_unlock_irqrestore(&ioapic_lock, flags);
                nr_ioapic_registers[apic] = reg_01.bits.entries+1;
        }
-       for (apic = 0; apic < nr_ioapics; apic++) {
+       for(apic = 0; apic < nr_ioapics; apic++) {
                int pin;
                /* See if any of the pins is in ExtINT mode */
                for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
                int pin;
                /* See if any of the pins is in ExtINT mode */
                for (pin = 0; pin < nr_ioapic_registers[apic]; pin++) {
@@ -1830,16 +1968,18 @@ void disable_IO_APIC(void)
                entry.dest_mode       = 0; /* Physical */
                entry.delivery_mode   = dest_ExtINT; /* ExtInt */
                entry.vector          = 0;
                entry.dest_mode       = 0; /* Physical */
                entry.delivery_mode   = dest_ExtINT; /* ExtInt */
                entry.vector          = 0;
-               entry.dest            = read_apic_id();
+               entry.dest            = read_apic_id();
 
                /*
                 * Add it to the IO-APIC irq-routing table:
                 */
                ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry);
        }
 
                /*
                 * Add it to the IO-APIC irq-routing table:
                 */
                ioapic_write_entry(ioapic_i8259.apic, ioapic_i8259.pin, entry);
        }
+
        disconnect_bsp_APIC(ioapic_i8259.pin != -1);
 }
 
        disconnect_bsp_APIC(ioapic_i8259.pin != -1);
 }
 
+#ifdef CONFIG_X86_32
 /*
  * function to set the IO-APIC physical IDs based on the
  * values stored in the MPC table.
 /*
  * function to set the IO-APIC physical IDs based on the
  * values stored in the MPC table.
@@ -1940,8 +2080,6 @@ static void __init setup_ioapic_ids_from_mpc(void)
 
                reg_00.bits.ID = mp_ioapics[apic].mp_apicid;
                spin_lock_irqsave(&ioapic_lock, flags);
 
                reg_00.bits.ID = mp_ioapics[apic].mp_apicid;
                spin_lock_irqsave(&ioapic_lock, flags);
-               io_apic_write(apic, 0, reg_00.raw);
-               spin_unlock_irqrestore(&ioapic_lock, flags);
 
                /*
                 * Sanity check
 
                /*
                 * Sanity check
@@ -1955,6 +2093,7 @@ static void __init setup_ioapic_ids_from_mpc(void)
                        apic_printk(APIC_VERBOSE, " ok.\n");
        }
 }
                        apic_printk(APIC_VERBOSE, " ok.\n");
        }
 }
+#endif
 
 int no_timer_check __initdata;
 
 
 int no_timer_check __initdata;
 
@@ -1994,9 +2133,10 @@ static int __init timer_irq_works(void)
         * might have cached one ExtINT interrupt.  Finally, at
         * least one tick may be lost due to delays.
         */
         * might have cached one ExtINT interrupt.  Finally, at
         * least one tick may be lost due to delays.
         */
+
+       /* jiffies wrap? */
        if (time_after(jiffies, t1 + 4))
                return 1;
        if (time_after(jiffies, t1 + 4))
                return 1;
-
        return 0;
 }
 
        return 0;
 }
 
@@ -2014,8 +2154,6 @@ static int __init timer_irq_works(void)
  */
 
 /*
  */
 
 /*
- * Startup quirk:
- *
  * Starting up a edge-triggered IO-APIC interrupt is
  * nasty - we need to make sure that we get the edge.
  * If it is already asserted for some reason, we need
  * Starting up a edge-triggered IO-APIC interrupt is
  * nasty - we need to make sure that we get the edge.
  * If it is already asserted for some reason, we need
@@ -2023,9 +2161,8 @@ static int __init timer_irq_works(void)
  *
  * This is not complete - we should be able to fake
  * an edge even if it isn't on the 8259A...
  *
  * This is not complete - we should be able to fake
  * an edge even if it isn't on the 8259A...
- *
- * (We do this for level-triggered IRQs too - it cannot hurt.)
  */
  */
+
 static unsigned int startup_ioapic_irq(unsigned int irq)
 {
        int was_pending = 0;
 static unsigned int startup_ioapic_irq(unsigned int irq)
 {
        int was_pending = 0;
@@ -2043,70 +2180,254 @@ static unsigned int startup_ioapic_irq(unsigned int irq)
        return was_pending;
 }
 
        return was_pending;
 }
 
+#ifdef CONFIG_X86_64
 static int ioapic_retrigger_irq(unsigned int irq)
 {
 static int ioapic_retrigger_irq(unsigned int irq)
 {
-       send_IPI_self(irq_cfg(irq)->vector);
+
+       struct irq_cfg *cfg = irq_cfg(irq);
+       unsigned long flags;
+
+       spin_lock_irqsave(&vector_lock, flags);
+       send_IPI_mask(cpumask_of_cpu(first_cpu(cfg->domain)), cfg->vector);
+       spin_unlock_irqrestore(&vector_lock, flags);
 
        return 1;
 }
 
        return 1;
 }
-
-#ifdef CONFIG_SMP
-asmlinkage void smp_irq_move_cleanup_interrupt(void)
+#else
+static int ioapic_retrigger_irq(unsigned int irq)
 {
 {
-       unsigned vector, me;
-       ack_APIC_irq();
-       irq_enter();
-
-       me = smp_processor_id();
-       for (vector = FIRST_EXTERNAL_VECTOR; vector < NR_VECTORS; vector++) {
-               unsigned int irq;
-               struct irq_desc *desc;
-               struct irq_cfg *cfg;
-               irq = __get_cpu_var(vector_irq)[vector];
+        send_IPI_self(irq_cfg(irq)->vector);
 
 
-               desc = irq_to_desc(irq);
-               if (!desc)
-                       continue;
+        return 1;
+}
+#endif
 
 
-               cfg = irq_cfg(irq);
-               spin_lock(&desc->lock);
-               if (!cfg->move_cleanup_count)
-                       goto unlock;
+/*
+ * Level and edge triggered IO-APIC interrupts need different handling,
+ * so we use two separate IRQ descriptors. Edge triggered IRQs can be
+ * handled with the level-triggered descriptor, but that one has slightly
+ * more overhead. Level-triggered interrupts cannot be handled with the
+ * edge-triggered handler, without risking IRQ storms and other ugly
+ * races.
+ */
 
 
-               if ((vector == cfg->vector) && cpu_isset(me, cfg->domain))
-                       goto unlock;
+#ifdef CONFIG_SMP
 
 
-               __get_cpu_var(vector_irq)[vector] = -1;
-               cfg->move_cleanup_count--;
-unlock:
-               spin_unlock(&desc->lock);
-       }
+#ifdef CONFIG_INTR_REMAP
+static void ir_irq_migration(struct work_struct *work);
 
 
-       irq_exit();
-}
+static DECLARE_DELAYED_WORK(ir_migration_work, ir_irq_migration);
 
 
-static void irq_complete_move(unsigned int irq)
+/*
+ * Migrate the IO-APIC irq in the presence of intr-remapping.
+ *
+ * For edge triggered, irq migration is a simple atomic update(of vector
+ * and cpu destination) of IRTE and flush the hardware cache.
+ *
+ * For level triggered, we need to modify the io-apic RTE aswell with the update
+ * vector information, along with modifying IRTE with vector and destination.
+ * So irq migration for level triggered is little  bit more complex compared to
+ * edge triggered migration. But the good news is, we use the same algorithm
+ * for level triggered migration as we have today, only difference being,
+ * we now initiate the irq migration from process context instead of the
+ * interrupt context.
+ *
+ * In future, when we do a directed EOI (combined with cpu EOI broadcast
+ * suppression) to the IO-APIC, level triggered irq migration will also be
+ * as simple as edge triggered migration and we can do the irq migration
+ * with a simple atomic update to IO-APIC RTE.
+ */
+static void migrate_ioapic_irq(int irq, cpumask_t mask)
 {
 {
-       struct irq_cfg *cfg = irq_cfg(irq);
-       unsigned vector, me;
+       struct irq_cfg *cfg;
+       struct irq_desc *desc;
+       cpumask_t tmp, cleanup_mask;
+       struct irte irte;
+       int modify_ioapic_rte;
+       unsigned int dest;
+       unsigned long flags;
 
 
-       if (likely(!cfg->move_in_progress))
+       cpus_and(tmp, mask, cpu_online_map);
+       if (cpus_empty(tmp))
                return;
 
                return;
 
-       vector = ~get_irq_regs()->orig_ax;
-       me = smp_processor_id();
-       if ((vector == cfg->vector) && cpu_isset(me, cfg->domain)) {
-               cpumask_t cleanup_mask;
+       if (get_irte(irq, &irte))
+               return;
 
 
-               cpus_and(cleanup_mask, cfg->old_domain, cpu_online_map);
-               cfg->move_cleanup_count = cpus_weight(cleanup_mask);
-               send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
+       if (assign_irq_vector(irq, mask))
+               return;
+
+       cfg = irq_cfg(irq);
+       cpus_and(tmp, cfg->domain, mask);
+       dest = cpu_mask_to_apicid(tmp);
+
+       desc = irq_to_desc(irq);
+       modify_ioapic_rte = desc->status & IRQ_LEVEL;
+       if (modify_ioapic_rte) {
+               spin_lock_irqsave(&ioapic_lock, flags);
+               __target_IO_APIC_irq(irq, dest, cfg->vector);
+               spin_unlock_irqrestore(&ioapic_lock, flags);
+       }
+
+       irte.vector = cfg->vector;
+       irte.dest_id = IRTE_DEST(dest);
+
+       /*
+        * Modified the IRTE and flushes the Interrupt entry cache.
+        */
+       modify_irte(irq, &irte);
+
+       if (cfg->move_in_progress) {
+               cpus_and(cleanup_mask, cfg->old_domain, cpu_online_map);
+               cfg->move_cleanup_count = cpus_weight(cleanup_mask);
+               send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
+               cfg->move_in_progress = 0;
+       }
+
+       desc->affinity = mask;
+}
+
+static int migrate_irq_remapped_level(int irq)
+{
+       int ret = -1;
+       struct irq_desc *desc = irq_to_desc(irq);
+
+       mask_IO_APIC_irq(irq);
+
+       if (io_apic_level_ack_pending(irq)) {
+               /*
+                * Interrupt in progress. Migrating irq now will change the
+                * vector information in the IO-APIC RTE and that will confuse
+                * the EOI broadcast performed by cpu.
+                * So, delay the irq migration to the next instance.
+                */
+               schedule_delayed_work(&ir_migration_work, 1);
+               goto unmask;
+       }
+
+       /* everthing is clear. we have right of way */
+       migrate_ioapic_irq(irq, desc->pending_mask);
+
+       ret = 0;
+       desc->status &= ~IRQ_MOVE_PENDING;
+       cpus_clear(desc->pending_mask);
+
+unmask:
+       unmask_IO_APIC_irq(irq);
+       return ret;
+}
+
+static void ir_irq_migration(struct work_struct *work)
+{
+       unsigned int irq;
+       struct irq_desc *desc;
+
+       for_each_irq_desc(irq, desc) {
+               if (desc->status & IRQ_MOVE_PENDING) {
+                       unsigned long flags;
+
+                       spin_lock_irqsave(&desc->lock, flags);
+                       if (!desc->chip->set_affinity ||
+                           !(desc->status & IRQ_MOVE_PENDING)) {
+                               desc->status &= ~IRQ_MOVE_PENDING;
+                               spin_unlock_irqrestore(&desc->lock, flags);
+                               continue;
+                       }
+
+                       desc->chip->set_affinity(irq, desc->pending_mask);
+                       spin_unlock_irqrestore(&desc->lock, flags);
+               }
+       }
+}
+
+/*
+ * Migrates the IRQ destination in the process context.
+ */
+static void set_ir_ioapic_affinity_irq(unsigned int irq, cpumask_t mask)
+{
+       struct irq_desc *desc = irq_to_desc(irq);
+
+       if (desc->status & IRQ_LEVEL) {
+               desc->status |= IRQ_MOVE_PENDING;
+               desc->pending_mask = mask;
+               migrate_irq_remapped_level(irq);
+               return;
+       }
+
+       migrate_ioapic_irq(irq, mask);
+}
+#endif
+
+asmlinkage void smp_irq_move_cleanup_interrupt(void)
+{
+       unsigned vector, me;
+       ack_APIC_irq();
+#ifdef CONFIG_X86_64
+       exit_idle();
+#endif
+       irq_enter();
+
+       me = smp_processor_id();
+       for (vector = FIRST_EXTERNAL_VECTOR; vector < NR_VECTORS; vector++) {
+               unsigned int irq;
+               struct irq_desc *desc;
+               struct irq_cfg *cfg;
+               irq = __get_cpu_var(vector_irq)[vector];
+
+               desc = irq_to_desc(irq);
+               if (!desc)
+                       continue;
+
+               cfg = irq_cfg(irq);
+               spin_lock(&desc->lock);
+               if (!cfg->move_cleanup_count)
+                       goto unlock;
+
+               if ((vector == cfg->vector) && cpu_isset(me, cfg->domain))
+                       goto unlock;
+
+               __get_cpu_var(vector_irq)[vector] = -1;
+               cfg->move_cleanup_count--;
+unlock:
+               spin_unlock(&desc->lock);
+       }
+
+       irq_exit();
+}
+
+static void irq_complete_move(unsigned int irq)
+{
+       struct irq_cfg *cfg = irq_cfg(irq);
+       unsigned vector, me;
+
+       if (likely(!cfg->move_in_progress))
+               return;
+
+       vector = ~get_irq_regs()->orig_ax;
+       me = smp_processor_id();
+       if ((vector == cfg->vector) && cpu_isset(me, cfg->domain)) {
+               cpumask_t cleanup_mask;
+
+               cpus_and(cleanup_mask, cfg->old_domain, cpu_online_map);
+               cfg->move_cleanup_count = cpus_weight(cleanup_mask);
+               send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
                cfg->move_in_progress = 0;
        }
 }
 #else
 static inline void irq_complete_move(unsigned int irq) {}
 #endif
                cfg->move_in_progress = 0;
        }
 }
 #else
 static inline void irq_complete_move(unsigned int irq) {}
 #endif
+#ifdef CONFIG_INTR_REMAP
+static void ack_x2apic_level(unsigned int irq)
+{
+       ack_x2APIC_irq();
+}
+
+static void ack_x2apic_edge(unsigned int irq)
+{
+       ack_x2APIC_irq();
+}
+#endif
 
 static void ack_apic_edge(unsigned int irq)
 {
 
 static void ack_apic_edge(unsigned int irq)
 {
@@ -2118,55 +2439,55 @@ static void ack_apic_edge(unsigned int irq)
 #ifdef CONFIG_X86_64
 static void ack_apic_level(unsigned int irq)
 {
 #ifdef CONFIG_X86_64
 static void ack_apic_level(unsigned int irq)
 {
-        int do_unmask_irq = 0;
+       int do_unmask_irq = 0;
 
 
-        irq_complete_move(irq);
+       irq_complete_move(irq);
 #ifdef CONFIG_GENERIC_PENDING_IRQ
 #ifdef CONFIG_GENERIC_PENDING_IRQ
-        /* If we are moving the irq we need to mask it */
-        if (unlikely(desc->status & IRQ_MOVE_PENDING)) {
-                do_unmask_irq = 1;
-                mask_IO_APIC_irq(irq);
-        }
+       /* If we are moving the irq we need to mask it */
+       if (unlikely(irq_to_desc(irq)->status & IRQ_MOVE_PENDING)) {
+               do_unmask_irq = 1;
+               mask_IO_APIC_irq(irq);
+       }
 #endif
 
 #endif
 
-        /*
-         * We must acknowledge the irq before we move it or the acknowledge will
-         * not propagate properly.
-         */
-        ack_APIC_irq();
-
-        /* Now we can move and renable the irq */
-        if (unlikely(do_unmask_irq)) {
-                /* Only migrate the irq if the ack has been received.
-                 *
-                 * On rare occasions the broadcast level triggered ack gets
-                 * delayed going to ioapics, and if we reprogram the
-                 * vector while Remote IRR is still set the irq will never
-                 * fire again.
-                 *
-                 * To prevent this scenario we read the Remote IRR bit
-                 * of the ioapic.  This has two effects.
-                 * - On any sane system the read of the ioapic will
-                 *   flush writes (and acks) going to the ioapic from
-                 *   this cpu.
-                 * - We get to see if the ACK has actually been delivered.
-                 *
-                 * Based on failed experiments of reprogramming the
-                 * ioapic entry from outside of irq context starting
-                 * with masking the ioapic entry and then polling until
-                 * Remote IRR was clear before reprogramming the
-                 * ioapic I don't trust the Remote IRR bit to be
-                 * completey accurate.
-                 *
-                 * However there appears to be no other way to plug
-                 * this race, so if the Remote IRR bit is not
-                 * accurate and is causing problems then it is a hardware bug
-                 * and you can go talk to the chipset vendor about it.
-                 */
-                if (!io_apic_level_ack_pending(irq))
-                        move_masked_irq(irq, desc);
-                unmask_IO_APIC_irq(irq);
-        }
+       /*
+        * We must acknowledge the irq before we move it or the acknowledge will
+        * not propagate properly.
+        */
+       ack_APIC_irq();
+
+       /* Now we can move and renable the irq */
+       if (unlikely(do_unmask_irq)) {
+               /* Only migrate the irq if the ack has been received.
+                *
+                * On rare occasions the broadcast level triggered ack gets
+                * delayed going to ioapics, and if we reprogram the
+                * vector while Remote IRR is still set the irq will never
+                * fire again.
+                *
+                * To prevent this scenario we read the Remote IRR bit
+                * of the ioapic.  This has two effects.
+                * - On any sane system the read of the ioapic will
+                *   flush writes (and acks) going to the ioapic from
+                *   this cpu.
+                * - We get to see if the ACK has actually been delivered.
+                *
+                * Based on failed experiments of reprogramming the
+                * ioapic entry from outside of irq context starting
+                * with masking the ioapic entry and then polling until
+                * Remote IRR was clear before reprogramming the
+                * ioapic I don't trust the Remote IRR bit to be
+                * completey accurate.
+                *
+                * However there appears to be no other way to plug
+                * this race, so if the Remote IRR bit is not
+                * accurate and is causing problems then it is a hardware bug
+                * and you can go talk to the chipset vendor about it.
+                */
+               if (!io_apic_level_ack_pending(irq))
+                       move_masked_irq(irq);
+               unmask_IO_APIC_irq(irq);
+       }
 }
 #else
 atomic_t irq_mis_count;
 }
 #else
 atomic_t irq_mis_count;
@@ -2177,25 +2498,25 @@ static void ack_apic_level(unsigned int irq)
 
        irq_complete_move(irq);
        move_native_irq(irq);
 
        irq_complete_move(irq);
        move_native_irq(irq);
-/*
- * It appears there is an erratum which affects at least version 0x11
- * of I/O APIC (that's the 82093AA and cores integrated into various
- * chipsets).  Under certain conditions a level-triggered interrupt is
- * erroneously delivered as edge-triggered one but the respective IRR
- * bit gets set nevertheless.  As a result the I/O unit expects an EOI
- * message but it will never arrive and further interrupts are blocked
- * from the source.  The exact reason is so far unknown, but the
- * phenomenon was observed when two consecutive interrupt requests
- * from a given source get delivered to the same CPU and the source is
- * temporarily disabled in between.
- *
- * A workaround is to simulate an EOI message manually.  We achieve it
- * by setting the trigger mode to edge and then to level when the edge
- * trigger mode gets detected in the TMR of a local APIC for a
- * level-triggered interrupt.  We mask the source for the time of the
- * operation to prevent an edge-triggered interrupt escaping meanwhile.
- * The idea is from Manfred Spraul.  --macro
- */
+       /*
      * It appears there is an erratum which affects at least version 0x11
      * of I/O APIC (that's the 82093AA and cores integrated into various
      * chipsets).  Under certain conditions a level-triggered interrupt is
      * erroneously delivered as edge-triggered one but the respective IRR
      * bit gets set nevertheless.  As a result the I/O unit expects an EOI
      * message but it will never arrive and further interrupts are blocked
      * from the source.  The exact reason is so far unknown, but the
      * phenomenon was observed when two consecutive interrupt requests
      * from a given source get delivered to the same CPU and the source is
      * temporarily disabled in between.
      *
      * A workaround is to simulate an EOI message manually.  We achieve it
      * by setting the trigger mode to edge and then to level when the edge
      * trigger mode gets detected in the TMR of a local APIC for a
      * level-triggered interrupt.  We mask the source for the time of the
      * operation to prevent an edge-triggered interrupt escaping meanwhile.
      * The idea is from Manfred Spraul.  --macro
      */
        i = irq_cfg(irq)->vector;
 
        v = apic_read(APIC_TMR + ((i & ~0x1f) >> 1));
        i = irq_cfg(irq)->vector;
 
        v = apic_read(APIC_TMR + ((i & ~0x1f) >> 1));
@@ -2225,6 +2546,20 @@ static struct irq_chip ioapic_chip __read_mostly = {
        .retrigger      = ioapic_retrigger_irq,
 };
 
        .retrigger      = ioapic_retrigger_irq,
 };
 
+#ifdef CONFIG_INTR_REMAP
+static struct irq_chip ir_ioapic_chip __read_mostly = {
+       .name           = "IR-IO-APIC",
+       .startup        = startup_ioapic_irq,
+       .mask           = mask_IO_APIC_irq,
+       .unmask         = unmask_IO_APIC_irq,
+       .ack            = ack_x2apic_edge,
+       .eoi            = ack_x2apic_level,
+#ifdef CONFIG_SMP
+       .set_affinity   = set_ir_ioapic_affinity_irq,
+#endif
+       .retrigger      = ioapic_retrigger_irq,
+};
+#endif
 
 static inline void init_IO_APIC_traps(void)
 {
 
 static inline void init_IO_APIC_traps(void)
 {
@@ -2282,7 +2617,7 @@ static void unmask_lapic_irq(unsigned int irq)
        apic_write(APIC_LVT0, v & ~APIC_LVT_MASKED);
 }
 
        apic_write(APIC_LVT0, v & ~APIC_LVT_MASKED);
 }
 
-static void ack_lapic_irq(unsigned int irq)
+static void ack_lapic_irq (unsigned int irq)
 {
        ack_APIC_irq();
 }
 {
        ack_APIC_irq();
 }
@@ -2383,12 +2718,12 @@ static inline void __init unlock_ExtINT_logic(void)
 
 static int disable_timer_pin_1 __initdata;
 /* Actually the next is obsolete, but keep it for paranoid reasons -AK */
 
 static int disable_timer_pin_1 __initdata;
 /* Actually the next is obsolete, but keep it for paranoid reasons -AK */
-static int __init parse_disable_timer_pin_1(char *arg)
+static int __init disable_timer_pin_setup(char *arg)
 {
        disable_timer_pin_1 = 1;
        return 0;
 }
 {
        disable_timer_pin_1 = 1;
        return 0;
 }
-early_param("disable_timer_pin_1", parse_disable_timer_pin_1);
+early_param("disable_timer_pin_1", disable_timer_pin_setup);
 
 int timer_through_8259 __initdata;
 
 
 int timer_through_8259 __initdata;
 
@@ -2397,6 +2732,8 @@ int timer_through_8259 __initdata;
  * a wide range of boards and BIOS bugs.  Fortunately only the timer IRQ
  * is so screwy.  Thanks to Brian Perkins for testing/hacking this beast
  * fanatically on his truly buggy board.
  * a wide range of boards and BIOS bugs.  Fortunately only the timer IRQ
  * is so screwy.  Thanks to Brian Perkins for testing/hacking this beast
  * fanatically on his truly buggy board.
+ *
+ * FIXME: really need to revamp this for all platforms.
  */
 static inline void __init check_timer(void)
 {
  */
 static inline void __init check_timer(void)
 {
@@ -2408,8 +2745,8 @@ static inline void __init check_timer(void)
 
        local_irq_save(flags);
 
 
        local_irq_save(flags);
 
-       ver = apic_read(APIC_LVR);
-       ver = GET_APIC_VERSION(ver);
+        ver = apic_read(APIC_LVR);
+        ver = GET_APIC_VERSION(ver);
 
        /*
         * get/set the timer IRQ vector:
 
        /*
         * get/set the timer IRQ vector:
@@ -2428,7 +2765,9 @@ static inline void __init check_timer(void)
         */
        apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT);
        init_8259A(1);
         */
        apic_write(APIC_LVT0, APIC_LVT_MASKED | APIC_DM_EXTINT);
        init_8259A(1);
+#ifdef CONFIG_X86_32
        timer_ack = (nmi_watchdog == NMI_IO_APIC && !APIC_INTEGRATED(ver));
        timer_ack = (nmi_watchdog == NMI_IO_APIC && !APIC_INTEGRATED(ver));
+#endif
 
        pin1  = find_isa_irq_pin(0, mp_INT);
        apic1 = find_isa_irq_apic(0, mp_INT);
 
        pin1  = find_isa_irq_pin(0, mp_INT);
        apic1 = find_isa_irq_apic(0, mp_INT);
@@ -2447,6 +2786,10 @@ static inline void __init check_timer(void)
         * 8259A.
         */
        if (pin1 == -1) {
         * 8259A.
         */
        if (pin1 == -1) {
+#ifdef CONFIG_INTR_REMAP
+               if (intr_remapping_enabled)
+                       panic("BIOS bug: timer not connected to IO-APIC");
+#endif
                pin1 = pin2;
                apic1 = apic2;
                no_pin1 = 1;
                pin1 = pin2;
                apic1 = apic2;
                no_pin1 = 1;
@@ -2473,6 +2816,10 @@ static inline void __init check_timer(void)
                                clear_IO_APIC_pin(0, pin1);
                        goto out;
                }
                                clear_IO_APIC_pin(0, pin1);
                        goto out;
                }
+#ifdef CONFIG_INTR_REMAP
+               if (intr_remapping_enabled)
+                       panic("timer doesn't work through Interrupt-remapped IO-APIC");
+#endif
                clear_IO_APIC_pin(apic1, pin1);
                if (!no_pin1)
                        apic_printk(APIC_QUIET, KERN_ERR "..MP-BIOS bug: "
                clear_IO_APIC_pin(apic1, pin1);
                if (!no_pin1)
                        apic_printk(APIC_QUIET, KERN_ERR "..MP-BIOS bug: "
@@ -2512,7 +2859,9 @@ static inline void __init check_timer(void)
                            "through the IO-APIC - disabling NMI Watchdog!\n");
                nmi_watchdog = NMI_NONE;
        }
                            "through the IO-APIC - disabling NMI Watchdog!\n");
                nmi_watchdog = NMI_NONE;
        }
+#ifdef CONFIG_X86_32
        timer_ack = 0;
        timer_ack = 0;
+#endif
 
        apic_printk(APIC_QUIET, KERN_INFO
                    "...trying to set up timer as Virtual Wire IRQ...\n");
 
        apic_printk(APIC_QUIET, KERN_INFO
                    "...trying to set up timer as Virtual Wire IRQ...\n");
@@ -2570,17 +2919,25 @@ out:
 
 void __init setup_IO_APIC(void)
 {
 
 void __init setup_IO_APIC(void)
 {
+
+#ifdef CONFIG_X86_32
        enable_IO_APIC();
        enable_IO_APIC();
+#else
+       /*
+        * calling enable_IO_APIC() is moved to setup_local_APIC for BP
+        */
+#endif
 
        io_apic_irqs = ~PIC_IRQS;
 
 
        io_apic_irqs = ~PIC_IRQS;
 
-       printk("ENABLING IO-APIC IRQs\n");
-
-       /*
-        * Set up IO-APIC IRQ routing.
-        */
-       if (!acpi_ioapic)
-               setup_ioapic_ids_from_mpc();
+       apic_printk(APIC_VERBOSE, "ENABLING IO-APIC IRQs\n");
+        /*
+         * Set up IO-APIC IRQ routing.
+         */
+#ifdef CONFIG_X86_32
+        if (!acpi_ioapic)
+                setup_ioapic_ids_from_mpc();
+#endif
        sync_Arb_IDs();
        setup_IO_APIC_irqs();
        init_IO_APIC_traps();
        sync_Arb_IDs();
        setup_IO_APIC_irqs();
        init_IO_APIC_traps();
@@ -2588,15 +2945,15 @@ void __init setup_IO_APIC(void)
 }
 
 /*
 }
 
 /*
- *     Called after all the initialization is done. If we didnt find any
- *     APIC bugs then we can allow the modify fast path
+ *      Called after all the initialization is done. If we didnt find any
+ *      APIC bugs then we can allow the modify fast path
  */
 
 static int __init io_apic_bug_finalize(void)
 {
  */
 
 static int __init io_apic_bug_finalize(void)
 {
-       if (sis_apic_bug == -1)
-               sis_apic_bug = 0;
-       return 0;
+        if (sis_apic_bug == -1)
+                sis_apic_bug = 0;
+        return 0;
 }
 
 late_initcall(io_apic_bug_finalize);
 }
 
 late_initcall(io_apic_bug_finalize);
@@ -2605,7 +2962,7 @@ struct sysfs_ioapic_data {
        struct sys_device dev;
        struct IO_APIC_route_entry entry[0];
 };
        struct sys_device dev;
        struct IO_APIC_route_entry entry[0];
 };
-static struct sysfs_ioapic_data *mp_ioapic_data[MAX_IO_APICS];
+static struct sysfs_ioapic_data * mp_ioapic_data[MAX_IO_APICS];
 
 static int ioapic_suspend(struct sys_device *dev, pm_message_t state)
 {
 
 static int ioapic_suspend(struct sys_device *dev, pm_message_t state)
 {
@@ -2615,8 +2972,8 @@ static int ioapic_suspend(struct sys_device *dev, pm_message_t state)
 
        data = container_of(dev, struct sysfs_ioapic_data, dev);
        entry = data->entry;
 
        data = container_of(dev, struct sysfs_ioapic_data, dev);
        entry = data->entry;
-       for (i = 0; i < nr_ioapic_registers[dev->id]; i++)
-               entry[i] = ioapic_read_entry(dev->id, i);
+       for (i = 0; i < nr_ioapic_registers[dev->id]; i ++, entry ++ )
+               *entry = ioapic_read_entry(dev->id, i);
 
        return 0;
 }
 
        return 0;
 }
@@ -2653,14 +3010,14 @@ static struct sysdev_class ioapic_sysdev_class = {
 
 static int __init ioapic_init_sysfs(void)
 {
 
 static int __init ioapic_init_sysfs(void)
 {
-       struct sys_device *dev;
-       int i, size, error = 0;
+       struct sys_device * dev;
+       int i, size, error;
 
        error = sysdev_class_register(&ioapic_sysdev_class);
        if (error)
                return error;
 
 
        error = sysdev_class_register(&ioapic_sysdev_class);
        if (error)
                return error;
 
-       for (i = 0; i < nr_ioapics; i++) {
+       for (i = 0; i < nr_ioapics; i++ ) {
                size = sizeof(struct sys_device) + nr_ioapic_registers[i]
                        * sizeof(struct IO_APIC_route_entry);
                mp_ioapic_data[i] = kzalloc(size, GFP_KERNEL);
                size = sizeof(struct sys_device) + nr_ioapic_registers[i]
                        * sizeof(struct IO_APIC_route_entry);
                mp_ioapic_data[i] = kzalloc(size, GFP_KERNEL);
@@ -2691,18 +3048,18 @@ device_initcall(ioapic_init_sysfs);
 unsigned int create_irq_nr(unsigned int irq_want)
 {
        /* Allocate an unused irq */
 unsigned int create_irq_nr(unsigned int irq_want)
 {
        /* Allocate an unused irq */
-       unsigned int irq, new;
+       unsigned int irq;
+       unsigned int new;
        unsigned long flags;
        struct irq_cfg *cfg_new;
 
 #ifndef CONFIG_HAVE_SPARSE_IRQ
        unsigned long flags;
        struct irq_cfg *cfg_new;
 
 #ifndef CONFIG_HAVE_SPARSE_IRQ
-       /* only can use bus/dev/fn.. when per_cpu vector is used */
        irq_want = nr_irqs - 1;
 #endif
 
        irq = 0;
        spin_lock_irqsave(&vector_lock, flags);
        irq_want = nr_irqs - 1;
 #endif
 
        irq = 0;
        spin_lock_irqsave(&vector_lock, flags);
-       for (new = (nr_irqs - 1); new > 0; new--) {
+       for (new = irq_want; new > 0; new--) {
                if (platform_legacy_irq(new))
                        continue;
                cfg_new = irq_cfg(new);
                if (platform_legacy_irq(new))
                        continue;
                cfg_new = irq_cfg(new);
@@ -2725,7 +3082,14 @@ unsigned int create_irq_nr(unsigned int irq_want)
 
 int create_irq(void)
 {
 
 int create_irq(void)
 {
-       return create_irq_nr(nr_irqs - 1);
+       int irq;
+
+       irq = create_irq_nr(nr_irqs - 1);
+
+       if (irq == 0)
+               irq = -1;
+
+       return irq;
 }
 
 void destroy_irq(unsigned int irq)
 }
 
 void destroy_irq(unsigned int irq)
@@ -2734,6 +3098,9 @@ void destroy_irq(unsigned int irq)
 
        dynamic_irq_cleanup(irq);
 
 
        dynamic_irq_cleanup(irq);
 
+#ifdef CONFIG_INTR_REMAP
+       free_irte(irq);
+#endif
        spin_lock_irqsave(&vector_lock, flags);
        __clear_irq_vector(irq);
        spin_unlock_irqrestore(&vector_lock, flags);
        spin_lock_irqsave(&vector_lock, flags);
        __clear_irq_vector(irq);
        spin_unlock_irqrestore(&vector_lock, flags);
@@ -2759,25 +3126,54 @@ static int msi_compose_msg(struct pci_dev *pdev, unsigned int irq, struct msi_ms
        cpus_and(tmp, cfg->domain, tmp);
        dest = cpu_mask_to_apicid(tmp);
 
        cpus_and(tmp, cfg->domain, tmp);
        dest = cpu_mask_to_apicid(tmp);
 
-       msg->address_hi = MSI_ADDR_BASE_HI;
-       msg->address_lo =
-               MSI_ADDR_BASE_LO |
-               ((INT_DEST_MODE == 0) ?
-                       MSI_ADDR_DEST_MODE_PHYSICAL:
-                       MSI_ADDR_DEST_MODE_LOGICAL) |
-               ((INT_DELIVERY_MODE != dest_LowestPrio) ?
-                       MSI_ADDR_REDIRECTION_CPU:
-                       MSI_ADDR_REDIRECTION_LOWPRI) |
-               MSI_ADDR_DEST_ID(dest);
-
-       msg->data =
-               MSI_DATA_TRIGGER_EDGE |
-               MSI_DATA_LEVEL_ASSERT |
-               ((INT_DELIVERY_MODE != dest_LowestPrio) ?
-                       MSI_DATA_DELIVERY_FIXED:
-                       MSI_DATA_DELIVERY_LOWPRI) |
-               MSI_DATA_VECTOR(cfg->vector);
+#ifdef CONFIG_INTR_REMAP
+       if (irq_remapped(irq)) {
+               struct irte irte;
+               int ir_index;
+               u16 sub_handle;
+
+               ir_index = map_irq_to_irte_handle(irq, &sub_handle);
+               BUG_ON(ir_index == -1);
+
+               memset (&irte, 0, sizeof(irte));
+
+               irte.present = 1;
+               irte.dst_mode = INT_DEST_MODE;
+               irte.trigger_mode = 0; /* edge */
+               irte.dlvry_mode = INT_DELIVERY_MODE;
+               irte.vector = cfg->vector;
+               irte.dest_id = IRTE_DEST(dest);
+
+               modify_irte(irq, &irte);
+
+               msg->address_hi = MSI_ADDR_BASE_HI;
+               msg->data = sub_handle;
+               msg->address_lo = MSI_ADDR_BASE_LO | MSI_ADDR_IR_EXT_INT |
+                                 MSI_ADDR_IR_SHV |
+                                 MSI_ADDR_IR_INDEX1(ir_index) |
+                                 MSI_ADDR_IR_INDEX2(ir_index);
+       } else
+#endif
+       {
+               msg->address_hi = MSI_ADDR_BASE_HI;
+               msg->address_lo =
+                       MSI_ADDR_BASE_LO |
+                       ((INT_DEST_MODE == 0) ?
+                               MSI_ADDR_DEST_MODE_PHYSICAL:
+                               MSI_ADDR_DEST_MODE_LOGICAL) |
+                       ((INT_DELIVERY_MODE != dest_LowestPrio) ?
+                               MSI_ADDR_REDIRECTION_CPU:
+                               MSI_ADDR_REDIRECTION_LOWPRI) |
+                       MSI_ADDR_DEST_ID(dest);
 
 
+               msg->data =
+                       MSI_DATA_TRIGGER_EDGE |
+                       MSI_DATA_LEVEL_ASSERT |
+                       ((INT_DELIVERY_MODE != dest_LowestPrio) ?
+                               MSI_DATA_DELIVERY_FIXED:
+                               MSI_DATA_DELIVERY_LOWPRI) |
+                       MSI_DATA_VECTOR(cfg->vector);
+       }
        return err;
 }
 
        return err;
 }
 
@@ -2788,6 +3184,7 @@ static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
        struct msi_msg msg;
        unsigned int dest;
        cpumask_t tmp;
        struct msi_msg msg;
        unsigned int dest;
        cpumask_t tmp;
+       struct irq_desc *desc;
 
        cpus_and(tmp, mask, cpu_online_map);
        if (cpus_empty(tmp))
 
        cpus_and(tmp, mask, cpu_online_map);
        if (cpus_empty(tmp))
@@ -2808,8 +3205,61 @@ static void set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
        msg.address_lo |= MSI_ADDR_DEST_ID(dest);
 
        write_msi_msg(irq, &msg);
        msg.address_lo |= MSI_ADDR_DEST_ID(dest);
 
        write_msi_msg(irq, &msg);
-       irq_to_desc(irq)->affinity = mask;
+       desc = irq_to_desc(irq);
+       desc->affinity = mask;
 }
 }
+
+#ifdef CONFIG_INTR_REMAP
+/*
+ * Migrate the MSI irq to another cpumask. This migration is
+ * done in the process context using interrupt-remapping hardware.
+ */
+static void ir_set_msi_irq_affinity(unsigned int irq, cpumask_t mask)
+{
+       struct irq_cfg *cfg;
+       unsigned int dest;
+       cpumask_t tmp, cleanup_mask;
+       struct irte irte;
+       struct irq_desc *desc;
+
+       cpus_and(tmp, mask, cpu_online_map);
+       if (cpus_empty(tmp))
+               return;
+
+       if (get_irte(irq, &irte))
+               return;
+
+       if (assign_irq_vector(irq, mask))
+               return;
+
+       cfg = irq_cfg(irq);
+       cpus_and(tmp, cfg->domain, mask);
+       dest = cpu_mask_to_apicid(tmp);
+
+       irte.vector = cfg->vector;
+       irte.dest_id = IRTE_DEST(dest);
+
+       /*
+        * atomically update the IRTE with the new destination and vector.
+        */
+       modify_irte(irq, &irte);
+
+       /*
+        * After this point, all the interrupts will start arriving
+        * at the new destination. So, time to cleanup the previous
+        * vector allocation.
+        */
+       if (cfg->move_in_progress) {
+               cpus_and(cleanup_mask, cfg->old_domain, cpu_online_map);
+               cfg->move_cleanup_count = cpus_weight(cleanup_mask);
+               send_IPI_mask(cleanup_mask, IRQ_MOVE_CLEANUP_VECTOR);
+               cfg->move_in_progress = 0;
+       }
+
+       desc = irq_to_desc(irq);
+       desc->affinity = mask;
+}
+#endif
 #endif /* CONFIG_SMP */
 
 /*
 #endif /* CONFIG_SMP */
 
 /*
@@ -2827,6 +3277,45 @@ static struct irq_chip msi_chip = {
        .retrigger      = ioapic_retrigger_irq,
 };
 
        .retrigger      = ioapic_retrigger_irq,
 };
 
+#ifdef CONFIG_INTR_REMAP
+static struct irq_chip msi_ir_chip = {
+       .name           = "IR-PCI-MSI",
+       .unmask         = unmask_msi_irq,
+       .mask           = mask_msi_irq,
+       .ack            = ack_x2apic_edge,
+#ifdef CONFIG_SMP
+       .set_affinity   = ir_set_msi_irq_affinity,
+#endif
+       .retrigger      = ioapic_retrigger_irq,
+};
+
+/*
+ * Map the PCI dev to the corresponding remapping hardware unit
+ * and allocate 'nvec' consecutive interrupt-remapping table entries
+ * in it.
+ */
+static int msi_alloc_irte(struct pci_dev *dev, int irq, int nvec)
+{
+       struct intel_iommu *iommu;
+       int index;
+
+       iommu = map_dev_to_ir(dev);
+       if (!iommu) {
+               printk(KERN_ERR
+                      "Unable to map PCI %s to iommu\n", pci_name(dev));
+               return -ENOENT;
+       }
+
+       index = alloc_irte(iommu, irq, nvec);
+       if (index < 0) {
+               printk(KERN_ERR
+                      "Unable to allocate %d IRTE for PCI %s\n", nvec,
+                       pci_name(dev));
+               return -ENOSPC;
+       }
+       return index;
+}
+#endif
 
 static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc, int irq)
 {
 
 static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc, int irq)
 {
@@ -2840,7 +3329,17 @@ static int setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc, int irq)
        set_irq_msi(irq, desc);
        write_msi_msg(irq, &msg);
 
        set_irq_msi(irq, desc);
        write_msi_msg(irq, &msg);
 
-       set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, "edge");
+#ifdef CONFIG_INTR_REMAP
+       if (irq_remapped(irq)) {
+               struct irq_desc *desc = irq_to_desc(irq);
+               /*
+                * irq migration in process context
+                */
+               desc->status |= IRQ_MOVE_PCNTXT;
+               set_irq_chip_and_handler_name(irq, &msi_ir_chip, handle_edge_irq, "edge");
+       } else
+#endif
+               set_irq_chip_and_handler_name(irq, &msi_chip, handle_edge_irq, "edge");
 
        return 0;
 }
 
        return 0;
 }
@@ -2859,59 +3358,164 @@ static unsigned int build_irq_for_pci_dev(struct pci_dev *dev)
 
 int arch_setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc)
 {
 
 int arch_setup_msi_irq(struct pci_dev *dev, struct msi_desc *desc)
 {
-       int irq, ret;
-
+       unsigned int irq;
+       int ret;
        unsigned int irq_want;
 
        irq_want = build_irq_for_pci_dev(dev) + 0x100;
 
        irq = create_irq_nr(irq_want);
        unsigned int irq_want;
 
        irq_want = build_irq_for_pci_dev(dev) + 0x100;
 
        irq = create_irq_nr(irq_want);
-
        if (irq == 0)
                return -1;
 
        if (irq == 0)
                return -1;
 
+#ifdef CONFIG_INTR_REMAP
+       if (!intr_remapping_enabled)
+               goto no_ir;
+
+       ret = msi_alloc_irte(dev, irq, 1);
+       if (ret < 0)
+               goto error;
+no_ir:
+#endif
        ret = setup_msi_irq(dev, desc, irq);
        if (ret < 0) {
                destroy_irq(irq);
                return ret;
        ret = setup_msi_irq(dev, desc, irq);
        if (ret < 0) {
                destroy_irq(irq);
                return ret;
-        }
-
+       }
        return 0;
        return 0;
+
+#ifdef CONFIG_INTR_REMAP
+error:
+       destroy_irq(irq);
+       return ret;
+#endif
 }
 
 int arch_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
 {
 }
 
 int arch_setup_msi_irqs(struct pci_dev *dev, int nvec, int type)
 {
-        unsigned int irq;
-        int ret, sub_handle;
-        struct msi_desc *desc;
-        unsigned int irq_want;
-
-        irq_want = build_irq_for_pci_dev(dev) + 0x100;
-        sub_handle = 0;
-        list_for_each_entry(desc, &dev->msi_list, list) {
-                irq = create_irq_nr(irq_want--);
-                if (irq == 0)
-                        return -1;
-                ret = setup_msi_irq(dev, desc, irq);
-                if (ret < 0)
-                        goto error;
-                sub_handle++;
-        }
-        return 0;
+       unsigned int irq;
+       int ret, sub_handle;
+       struct msi_desc *desc;
+       unsigned int irq_want;
+
+#ifdef CONFIG_INTR_REMAP
+       struct intel_iommu *iommu = 0;
+       int index = 0;
+#endif
+
+       irq_want = build_irq_for_pci_dev(dev) + 0x100;
+       sub_handle = 0;
+       list_for_each_entry(desc, &dev->msi_list, list) {
+               irq = create_irq_nr(irq_want--);
+               if (irq == 0)
+                       return -1;
+#ifdef CONFIG_INTR_REMAP
+               if (!intr_remapping_enabled)
+                       goto no_ir;
+
+               if (!sub_handle) {
+                       /*
+                        * allocate the consecutive block of IRTE's
+                        * for 'nvec'
+                        */
+                       index = msi_alloc_irte(dev, irq, nvec);
+                       if (index < 0) {
+                               ret = index;
+                               goto error;
+                       }
+               } else {
+                       iommu = map_dev_to_ir(dev);
+                       if (!iommu) {
+                               ret = -ENOENT;
+                               goto error;
+                       }
+                       /*
+                        * setup the mapping between the irq and the IRTE
+                        * base index, the sub_handle pointing to the
+                        * appropriate interrupt remap table entry.
+                        */
+                       set_irte_irq(irq, iommu, index, sub_handle);
+               }
+no_ir:
+#endif
+               ret = setup_msi_irq(dev, desc, irq);
+               if (ret < 0)
+                       goto error;
+               sub_handle++;
+       }
+       return 0;
 
 error:
 
 error:
-        destroy_irq(irq);
-        return ret;
+       destroy_irq(irq);
+       return ret;
 }
 
 }
 
-
 void arch_teardown_msi_irq(unsigned int irq)
 {
        destroy_irq(irq);
 }
 
 void arch_teardown_msi_irq(unsigned int irq)
 {
        destroy_irq(irq);
 }
 
-#endif /* CONFIG_PCI_MSI */
+#ifdef CONFIG_DMAR
+#ifdef CONFIG_SMP
+static void dmar_msi_set_affinity(unsigned int irq, cpumask_t mask)
+{
+       struct irq_cfg *cfg;
+       struct msi_msg msg;
+       unsigned int dest;
+       cpumask_t tmp;
+       struct irq_desc *desc;
+
+       cpus_and(tmp, mask, cpu_online_map);
+       if (cpus_empty(tmp))
+               return;
+
+       if (assign_irq_vector(irq, mask))
+               return;
+
+       cfg = irq_cfg(irq);
+       cpus_and(tmp, cfg->domain, mask);
+       dest = cpu_mask_to_apicid(tmp);
+
+       dmar_msi_read(irq, &msg);
+
+       msg.data &= ~MSI_DATA_VECTOR_MASK;
+       msg.data |= MSI_DATA_VECTOR(cfg->vector);
+       msg.address_lo &= ~MSI_ADDR_DEST_ID_MASK;
+       msg.address_lo |= MSI_ADDR_DEST_ID(dest);
+
+       dmar_msi_write(irq, &msg);
+       desc = irq_to_desc(irq);
+       desc->affinity = mask;
+}
+#endif /* CONFIG_SMP */
+
+struct irq_chip dmar_msi_type = {
+       .name = "DMAR_MSI",
+       .unmask = dmar_msi_unmask,
+       .mask = dmar_msi_mask,
+       .ack = ack_apic_edge,
+#ifdef CONFIG_SMP
+       .set_affinity = dmar_msi_set_affinity,
+#endif
+       .retrigger = ioapic_retrigger_irq,
+};
+
+int arch_setup_dmar_msi(unsigned int irq)
+{
+       int ret;
+       struct msi_msg msg;
 
 
+       ret = msi_compose_msg(NULL, irq, &msg);
+       if (ret < 0)
+               return ret;
+       dmar_msi_write(irq, &msg);
+       set_irq_chip_and_handler_name(irq, &dmar_msi_type, handle_edge_irq,
+               "edge");
+       return 0;
+}
+#endif
+
+#endif /* CONFIG_PCI_MSI */
 /*
  * Hypertransport interrupt support
  */
 /*
  * Hypertransport interrupt support
  */
@@ -2938,6 +3542,7 @@ static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask)
        struct irq_cfg *cfg;
        unsigned int dest;
        cpumask_t tmp;
        struct irq_cfg *cfg;
        unsigned int dest;
        cpumask_t tmp;
+       struct irq_desc *desc;
 
        cpus_and(tmp, mask, cpu_online_map);
        if (cpus_empty(tmp))
 
        cpus_and(tmp, mask, cpu_online_map);
        if (cpus_empty(tmp))
@@ -2951,7 +3556,8 @@ static void set_ht_irq_affinity(unsigned int irq, cpumask_t mask)
        dest = cpu_mask_to_apicid(tmp);
 
        target_ht_irq(irq, dest, cfg->vector);
        dest = cpu_mask_to_apicid(tmp);
 
        target_ht_irq(irq, dest, cfg->vector);
-       irq_to_desc(irq)->affinity = mask;
+       desc = irq_to_desc(irq);
+       desc->affinity = mask;
 }
 #endif
 
 }
 #endif
 
@@ -2974,7 +3580,7 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
 
        tmp = TARGET_CPUS;
        err = assign_irq_vector(irq, tmp);
 
        tmp = TARGET_CPUS;
        err = assign_irq_vector(irq, tmp);
-       if ( !err) {
+       if (!err) {
                struct ht_irq_msg msg;
                unsigned dest;
 
                struct ht_irq_msg msg;
                unsigned dest;
 
@@ -3007,11 +3613,12 @@ int arch_setup_ht_irq(unsigned int irq, struct pci_dev *dev)
 #endif /* CONFIG_HT_IRQ */
 
 /* --------------------------------------------------------------------------
 #endif /* CONFIG_HT_IRQ */
 
 /* --------------------------------------------------------------------------
-                       ACPI-based IOAPIC Configuration
+                          ACPI-based IOAPIC Configuration
    -------------------------------------------------------------------------- */
 
 #ifdef CONFIG_ACPI
 
    -------------------------------------------------------------------------- */
 
 #ifdef CONFIG_ACPI
 
+#ifdef CONFIG_X86_32
 int __init io_apic_get_unique_id(int ioapic, int apic_id)
 {
        union IO_APIC_reg_00 reg_00;
 int __init io_apic_get_unique_id(int ioapic, int apic_id)
 {
        union IO_APIC_reg_00 reg_00;
@@ -3086,7 +3693,6 @@ int __init io_apic_get_unique_id(int ioapic, int apic_id)
        return apic_id;
 }
 
        return apic_id;
 }
 
-
 int __init io_apic_get_version(int ioapic)
 {
        union IO_APIC_reg_01    reg_01;
 int __init io_apic_get_version(int ioapic)
 {
        union IO_APIC_reg_01    reg_01;
@@ -3098,9 +3704,9 @@ int __init io_apic_get_version(int ioapic)
 
        return reg_01.bits.version;
 }
 
        return reg_01.bits.version;
 }
+#endif
 
 
-
-int __init io_apic_get_redir_entries(int ioapic)
+int __init io_apic_get_redir_entries (int ioapic)
 {
        union IO_APIC_reg_01    reg_01;
        unsigned long flags;
 {
        union IO_APIC_reg_01    reg_01;
        unsigned long flags;
@@ -3113,10 +3719,10 @@ int __init io_apic_get_redir_entries(int ioapic)
 }
 
 
 }
 
 
-int io_apic_set_pci_routing(int ioapic, int pin, int irq, int triggering, int polarity)
+int io_apic_set_pci_routing (int ioapic, int pin, int irq, int triggering, int polarity)
 {
        if (!IO_APIC_IRQ(irq)) {
 {
        if (!IO_APIC_IRQ(irq)) {
-               printk(KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
+               apic_printk(APIC_QUIET,KERN_ERR "IOAPIC[%d]: Invalid reference to IRQ 0\n",
                        ioapic);
                return -EINVAL;
        }
                        ioapic);
                return -EINVAL;
        }
@@ -3132,6 +3738,7 @@ int io_apic_set_pci_routing(int ioapic, int pin, int irq, int triggering, int po
        return 0;
 }
 
        return 0;
 }
 
+
 int acpi_get_override_irq(int bus_irq, int *trigger, int *polarity)
 {
        int i;
 int acpi_get_override_irq(int bus_irq, int *trigger, int *polarity)
 {
        int i;
@@ -3163,7 +3770,6 @@ void __init setup_ioapic_dest(void)
 {
        int pin, ioapic, irq, irq_entry;
        struct irq_cfg *cfg;
 {
        int pin, ioapic, irq, irq_entry;
        struct irq_cfg *cfg;
-       struct irq_desc *desc;
 
        if (skip_ioapic_setup == 1)
                return;
 
        if (skip_ioapic_setup == 1)
                return;
@@ -3184,43 +3790,124 @@ void __init setup_ioapic_dest(void)
                                setup_IO_APIC_irq(ioapic, pin, irq,
                                                  irq_trigger(irq_entry),
                                                  irq_polarity(irq_entry));
                                setup_IO_APIC_irq(ioapic, pin, irq,
                                                  irq_trigger(irq_entry),
                                                  irq_polarity(irq_entry));
-                       else {
-                               desc = irq_to_desc(irq);
+#ifdef CONFIG_INTR_REMAP
+                       else if (intr_remapping_enabled)
+                               set_ir_ioapic_affinity_irq(irq, TARGET_CPUS);
+#endif
+                       else
                                set_ioapic_affinity_irq(irq, TARGET_CPUS);
                                set_ioapic_affinity_irq(irq, TARGET_CPUS);
-                       }
                }
 
        }
 }
 #endif
 
                }
 
        }
 }
 #endif
 
+#ifdef CONFIG_X86_64
+#define IOAPIC_RESOURCE_NAME_SIZE 11
+
+static struct resource *ioapic_resources;
+
+static struct resource * __init ioapic_setup_resources(void)
+{
+       unsigned long n;
+       struct resource *res;
+       char *mem;
+       int i;
+
+       if (nr_ioapics <= 0)
+               return NULL;
+
+       n = IOAPIC_RESOURCE_NAME_SIZE + sizeof(struct resource);
+       n *= nr_ioapics;
+
+       mem = alloc_bootmem(n);
+       res = (void *)mem;
+
+       if (mem != NULL) {
+               mem += sizeof(struct resource) * nr_ioapics;
+
+               for (i = 0; i < nr_ioapics; i++) {
+                       res[i].name = mem;
+                       res[i].flags = IORESOURCE_MEM | IORESOURCE_BUSY;
+                       sprintf(mem,  "IOAPIC %u", i);
+                       mem += IOAPIC_RESOURCE_NAME_SIZE;
+               }
+       }
+
+       ioapic_resources = res;
+
+       return res;
+}
+#endif
+
 void __init ioapic_init_mappings(void)
 {
        unsigned long ioapic_phys, idx = FIX_IO_APIC_BASE_0;
        int i;
 void __init ioapic_init_mappings(void)
 {
        unsigned long ioapic_phys, idx = FIX_IO_APIC_BASE_0;
        int i;
+#ifdef CONFIG_X86_64
+       struct resource *ioapic_res;
 
 
+       ioapic_res = ioapic_setup_resources();
+#endif
        for (i = 0; i < nr_ioapics; i++) {
                if (smp_found_config) {
                        ioapic_phys = mp_ioapics[i].mp_apicaddr;
        for (i = 0; i < nr_ioapics; i++) {
                if (smp_found_config) {
                        ioapic_phys = mp_ioapics[i].mp_apicaddr;
-                       if (!ioapic_phys) {
-                               printk(KERN_ERR
-                                      "WARNING: bogus zero IO-APIC "
-                                      "address found in MPTABLE, "
-                                      "disabling IO/APIC support!\n");
-                               smp_found_config = 0;
-                               skip_ioapic_setup = 1;
-                               goto fake_ioapic_page;
-                       }
+#ifdef CONFIG_X86_32
+                        if (!ioapic_phys) {
+                                printk(KERN_ERR
+                                       "WARNING: bogus zero IO-APIC "
+                                       "address found in MPTABLE, "
+                                       "disabling IO/APIC support!\n");
+                                smp_found_config = 0;
+                                skip_ioapic_setup = 1;
+                                goto fake_ioapic_page;
+                        }
+#endif
                } else {
                } else {
+#ifdef CONFIG_X86_32
 fake_ioapic_page:
 fake_ioapic_page:
+#endif
                        ioapic_phys = (unsigned long)
                        ioapic_phys = (unsigned long)
-                                     alloc_bootmem_pages(PAGE_SIZE);
+                               alloc_bootmem_pages(PAGE_SIZE);
                        ioapic_phys = __pa(ioapic_phys);
                }
                set_fixmap_nocache(idx, ioapic_phys);
                        ioapic_phys = __pa(ioapic_phys);
                }
                set_fixmap_nocache(idx, ioapic_phys);
-               printk(KERN_DEBUG "mapped IOAPIC to %08lx (%08lx)\n",
-                      __fix_to_virt(idx), ioapic_phys);
+               apic_printk(APIC_VERBOSE,
+                           "mapped IOAPIC to %08lx (%08lx)\n",
+                           __fix_to_virt(idx), ioapic_phys);
                idx++;
                idx++;
+
+#ifdef CONFIG_X86_64
+               if (ioapic_res != NULL) {
+                       ioapic_res->start = ioapic_phys;
+                       ioapic_res->end = ioapic_phys + (4 * 1024) - 1;
+                       ioapic_res++;
+               }
+#endif
        }
 }
 
        }
 }
 
+#ifdef CONFIG_X86_64
+static int __init ioapic_insert_resources(void)
+{
+       int i;
+       struct resource *r = ioapic_resources;
+
+       if (!r) {
+               printk(KERN_ERR
+                      "IO APIC resources could be not be allocated.\n");
+               return -1;
+       }
+
+       for (i = 0; i < nr_ioapics; i++) {
+               insert_resource(&iomem_resource, r);
+               r++;
+       }
+
+       return 0;
+}
+
+/* Insert the IO APIC resources after PCI initialization has occured to handle
+ * IO APICS that are mapped in on a BAR in PCI space. */
+late_initcall(ioapic_insert_resources);
+#endif